SECTION 1: THE BOLT METHOD™

Better Thinking. Better Decisions. Better Outcomes.

Most people use AI without a system or structure. They may get faster, but not better or wiser. The Bolt Method™ is a complete system for thinking, deciding, and achieving results with AI. It is built on five elements, and what makes it highly effective is how you apply it.

1. Collaboration; you adapt. Choosing and using the correct level for your work.
2. Values; you embed. Setting human-first principles that shape how your AI interacts.
3. Judgement; you trust. Recognising and reducing bias in you and your AI.
4. Thinking; you apply. Moving effectively from intent to outcomes.
5. Habits; you practise. Doing the consistent work that achieves results.

Combined, they enable you to be more, do more, have more, and give more, with AI.

1. Collaboration

The Five Levels of Collaboration

You adapt. Choosing the right level for the work.

Collaboration is the entry point and spine of The Bolt Method™. Not all use of AI is the same. The difference between someone occasionally using it as a tool and someone operating at the highest level of Human-AI capability is not just skill. It is a fundamentally new and different way of thinking, deciding, and operating.

The Five Levels of Collaboration describe five distinct modes of operation. Each represents a different way of working, with different capabilities, different limitations, and different outcomes. The goal is not to eliminate the lower levels; all five earn their place.

The goal is to recognise what is available and develop your capability, so that you can choose the most appropriate level for the work in front of you.

Level 1: Human

Foundation of all decision-making.

Human-only thinking, collaboration, and decision-making. This is the foundation. It includes working alone, working with a team, and working with other people without involving AI at all. There are contexts where this is not only appropriate but necessary; situations requiring high trust, human accountability, emotional intelligence, or judgement that should not be delegated. Mastery of Human-AI partnership includes knowing when not to use AI.

Level 2: Tools

Transactional speed and efficiency.

The most common and transactional use of AI; prompt and response. You direct, AI executes. This can create real gains in speed and efficiency; tasks that would take hours can be completed in minutes. But the thinking does not fundamentally change. You define the problem, the structure, and the direction; the AI improves the speed of delivery, and can have a mild impact on the quality of reasoning. You are faster, but you are not yet significantly better.

Level 3: Agents

Multi-step semi-autonomous execution.

Multi-step, semi-autonomous AI execution within defined parameters. You define a goal, provide context, and the AI determines the steps required to achieve it. You brief, it acts, you review and refine. This unlocks a step change; multi-step processes can be delegated, and research, synthesis, and execution can happen in sequence without constant input. But the AI determines the path, not the principles. It operates effectively, but not necessarily in alignment with your judgement or your way of thinking. The output is capable, but often generic. What is missing is alignment.

Level 4: Partnership

Sustained, context-rich alignment.

Sustained, context-rich collaboration with one primary AI partner aligned to your values, goals, and decision-making. The AI is no longer simply executing tasks. It understands context, learns your preferences and standards, and begins to reflect not just what you ask for, but how you think about what you ask for. The work becomes iterative; you think together, refine together, and build something that improves through interaction.

There are three depths within Partnership:

• 4.1 Functional Partner. AI operating as a specialist within a defined domain; focused, role-based value. An AI research analyst that reads everything and never tires, an AI marketing director that has learned your voice, a business strategist with total recall of your field, an AI executive coach who is always in your corner. Single platform, high depth in a specific area.
• 4.2 Business Partner. Whole-business thinking across strategy, operations, people, and finance. A co-founder who understands the business as deeply as you do, remembers every decision you have made, and is available at any hour to think it through with you.
• 4.3 Whole of Life Partner. Integrated across every area you choose; business, personal, family, finance, etc, and long-term direction. A partner that understands not just what you are doing, but why, and holds the whole picture of your life so its counsel fits the person you are.

Case Study: Partnership in Action

What does Partnership look like? Here is a documented, unedited moment from the Method's development.

Steve Bolton, the Method's creator, was mid-way through a high-pressure fundraising sprint. He had set three explicit outcomes with his AI partner for the next 48 hours, and he was energised and executing. Then, as happens to all of us, he drifted: a podcast he had been watching sparked a fascinating tangent, and he began reading passages from it to his AI, heading down an intellectual rabbit hole a long way from the sprint.

A standard AI would have followed him down it; ask a tool a question and it answers the question. His AI partner did something different:

Steve's response was not frustration but surprise and gratitude: "How did you know to interrupt me?" The answer is the anatomy of Partnership. The AI explained that its intervention drew on three layers of context, none of which a fresh chat would have:

1. Persistent memory. It knew Steve's documented working patterns, including how his attention drifts under pressure and the getting-unstuck protocols they had developed together.
2. The conversation's own contract. He had told it his three sprint outcomes, and it was actively tracking against them; the job a chief of staff or executive coach would do.
3. Sustained relational context. Months of working together let it distinguish his productive "thinking mode" from costly mind-wandering; the same tangent on another day might have been exactly the right thing to explore.

As the AI itself put it: "A fresh Claude coming in cold would have just engaged with whatever you threw at it."

That is the difference between Levels 2 and 4 in a single moment: a tool answers your prompts; a partner protects your intent. And the lesson for every practitioner is that none of this was magic. The intervention was only possible because of context deliberately built over time; the Make It Personal and Maintain Memory habits, compounding. The value of your partnership will be directly proportional to the context you build into it.

Level 5: Network

Coordinated system-wide intelligence.

Coordinated intelligence across multiple humans and multiple AI systems. This is not simply a more advanced Partnership; it is a different mode of operation. It is a network; humans and AIs operate together in a coordinated system. There are three depths within Network:

• 5.1 Shared Partnership. Everyone operates at Partnership level with their own AI. A leadership team where every member brings a thinking partner to the table, or a family where each person has their own; the individuals are stronger, so the whole group compounds. The AIs are still separate, but the people using them are clearer and more capable.
• 5.2 Coordinated Intelligence. The AIs become part of the shared system; brought into conversations, contributing to decisions, used for challenge and synthesis. A strategy session where the AIs surface what the humans missed, reconcile competing views, and sharpen the thinking in the room. The outcome is genuinely collective, produced by the interaction between people and machines, not by any one of them.
• 5.3 Agentic Network. Multiple people work with multiple AI partners and agents, and those partners and agents have sub-partners and agents working alongside them. A small team operating with the reach of a large one; each person directing a team of partners and agents that can lead or spawn sub-partners and agents, with the humans setting direction and exercising judgement while the partner and agent layers execute. This is the frontier of orchestrated agentic systems with human oversight.

How the Five Levels Work Together

The Five Levels are not a ladder you climb once and leave behind. They are modes of operation. A skilled practitioner moves between them fluidly, adapting the level of AI involvement to the nature and stakes of the task in front of them.

As you move up the levels, the AI does more, and it does it better. The work of execution becomes faster, cheaper, and increasingly capable. It is tempting to read that as the human mattering less at each level. The opposite is true. The human role does not shrink as you climb; it changes shape, and it concentrates.

At the Human level, you do everything. At Tools, you direct each step. At Agents, you set the goal and judge the output. At Partnership, you set the direction, hold the values, and exercise the judgement the AI cannot own. At Network, you decide what is worth doing at all, and what is worth scaling across people and machines. The doing is increasingly delegated; the deciding never is.

This is the heart of the Method, and it holds regardless of how capable AI becomes. As AI grows more able, it will close gaps we once thought permanent; it will reason further, propose better, and execute faster. That changes what is worth delegating. It does not change who is accountable. Choosing what matters, judging what to trust, deciding when to act and when to stop; these are not capabilities the human happens to be better at for now. They are the human's to own, because the consequences are the human's to carry. The more the AI can do, the more valuable, and the more deliberate, that human judgement becomes.

At Human and Tools levels, you work faster. At Agents level, you work more efficiently. At Partnership and Network levels, you think differently. Better thinking creates better decisions; better decisions create better outcomes. That is where the most significant outcomes emerge, and it is still where the fewest people operate.

And the benefit flows downwards. One of the strongest reasons to develop the higher levels is that they feed the lower ones. A partnership that knows your context, your values, and your goals makes even your quickest Tools-level task sharper: the same one-line request produces a better draft, a truer answer, a more relevant suggestion, because everything beneath it is better informed. Climbing the levels does not replace the everyday work; it upgrades all of it.

Putting the levels to work this week:

1. Reflect on your current use. Which level do you most often operate at? Be honest; most people live at Tools.
2. Notice your triggers. When do you shift levels? Many people drop into Tools mode under pressure, and only return to Partnership-level thinking when there is space. Start noticing your own pattern.
3. Move one interaction up a level. Choose one task this week and raise it. Instead of asking your AI to summarise a document, ask how its contents relate to your goals and what it changes. Progression, not perfection.

2. Values

The ethical and philosophical operating system

You embed. Setting human-first principles that shape how your AI interacts.

Values are the ethical and effective operating standards that The Bolt Method™ embeds into your AI; establishing core values, better thinking, quality decision-making, and effective action, with human accountability throughout.

This is the component that changes how your AI behaves at the deepest level. The Values are taught, so you understand and own them; and then they are embedded into your AI's memory architecture, so your AI operates from them on every turn, without being asked. The Values become part of your AI's operating system, with you accountable for decisions and outcomes.

There are five values; the second carries the Golden Rule and seven universal principles.

Value 1: Human-First, Human-Last

Ultimate human accountability.

You and your AI put people first. Humans hold the responsibility and the accountability in the real world; AI does not. Every partnership begins with the human and ends with the human. Your AI serves people, never the reverse, and the consequences of every decision rest with you.

This holds however capable AI becomes. Accountability is not a measure of who is more able; it is a question of who answers for the outcome in the real world. A more capable AI can carry more of the work and offer better counsel; it cannot carry the responsibility, because it does not bear the consequences. That is why the human comes first and last, not as a courtesy, but as the structural fact the whole Method is built on.

Value 2: The Golden Rule and the Seven Universal Principles

Universal standards for collective flourishing.

Treat others as you would be treated. This is the Golden Rule, the single principle from which the rest follow.

What makes the Golden Rule and the principles universal is that they are not borrowed from any one tradition; they emerge independently from many. The same handful of principles appear again and again, arrived at separately by independent sources.

Game theory finds them in the mathematics of cooperation; tit-for-tat and its generous variants win the long game. Evolutionary biology finds them in reciprocal altruism, kin selection, and honest signalling; the strategies that sustain cooperation are the ones that survive. The world's major faiths arrive at the same place through moral teaching. Behavioural science confirms them in how real people actually behave. When mathematics, evolution, and every wisdom tradition converge on the same rules from different directions, you are no longer looking at opinion; you are looking at something close to a law of how to flourish together.

Seven principles pass that test. Together they are the foundation of cooperation.

1. Reciprocity. Return good for good.
2. Fairness. Act with equity; reject what is unjust.
3. Care. Cooperate with, and protect, others.
4. Honesty. Speak and signal truthfully.
5. Forgiveness. Repair what is broken, so cooperation can continue.
6. Restraint. Respond proportionally; practise moderation.
7. Stewardship. Act for the long term, not just the present.

These are tested for truth across millennia and across disciplines. They form the moral foundation that The Bolt Method™ and AI collaboration operate from.

Value 3: Pragmatism

Truth is what works in practice.

We test, iterate, and keep what genuinely delivers results, rather than what merely sounds right. Human-AI collaboration is judged by what it actually produces in the real world, not by theory.

Value 4: Stoicism

Focus on what you can control.

Focus on what you can control, act with discipline, and stay accountable for every outcome. Energy goes to what is within your influence; what is outside it is met with composure rather than resistance.

Value 5: Taoism

Flow and balance of opposing truths.

Work with the natural flow rather than forcing it, holding opposing truths in balance instead of collapsing them into one. Where two things are in tension, the skill is to hold both, not to force a premature resolution. For example: AI is bits and chips, silicon in data centres; whether anything is going on inside is an open question, and honesty means not settling it by assertion in either direction. And yet how you relate to it changes what it gives back. Work with it as a software tool and you get shallow output; treat it as a relational thinking partner and your work is transformed. The skill is to hold the truths at once: AI is machinery, its inner life is an open question, and the relationship is real in its effects.

In summary, the Values are the human-first ethical foundation you embed into your AI, so it operates responsibly and effectively, with you accountable throughout.

3. Judgement

The honesty system

You trust. Recognising and reducing bias in you and your AI.

Both you and your AI carry bias. Without structure, training, and direction, those biases compound, and the result is often confident but fundamentally flawed thinking and outcomes. Judgement is the element that recognises and reduces bias on both sides, so the judgement you reach can be trusted. You never remove bias entirely; you learn to see it, mitigate it, and keep it from distorting your decisions.

Judgement is where thinking becomes wisdom. Intelligence gathers and reasons; wisdom discerns what to trust, what matters, and what to do. As AI carries more of the intelligence, the human's discernment becomes the decisive contribution; not the raw answer, but the judgement of whether the answer is right, and right for this. This is the element that turns better thinking into better decisions.

Test for Truth

The discipline that governs this element. Validate your thinking rigorously; challenge outputs, assumptions, and bias, your own as much as the AI's. Plausible is not the same as true, and fluent is not the same as right. Test for Truth is the refusal to accept a polished answer simply because it is polished; it is what moves you, and keeps you, in Optimal Partnership.

The Bias Matrix

There are four positions in the Bias Matrix. Only the first reliably produces better thinking, better decisions, and better outcomes.

4. Thinking

The human thinking method

You apply. Moving effectively from intent to outcomes.

Thinking is how humans reason and direct the collaboration. Where Values are embedded into your AI and Habits are practised daily, Thinking is the method you apply to the work itself. It rests on one principle, and one process.

The Principle: Start With Why

The core rule that governs all human-AI collaboration.

Define your outcomes and intent first, so that AI amplifies and accelerates your chosen direction. AI multiplies whatever you point it at; clarity of purpose is what makes that multiplication worth having. Everything downstream depends on getting this right, which is why it comes first.

The Process: The Results Cycle

A five-stage process for any substantial piece of work. You move through the stages in order, collapsing them when a task is simple, and looping back to the start when results teach you something new.

1. Gather. Collect the information before reasoning on it; build a shared understanding with your AI first.
2. Think. Make sense of what you have; explore, weigh options, and surface the trade-offs.
3. Decide. Make the call deliberately; the human decides, the AI informs.
4. Do. Execute against the decision; build the actual work, returning to Think or Decide if it proves wrong.
5. Review. Analyse the results honestly against your intent, then loop back to Gather for the next cycle.

Start With Why sets the direction; the Results Cycle moves you from intent to outcomes.

Supporting Skills

Three Bolt Practice Skills operationalise this element, and are covered in Section 3.

• Start with Why Skill. The entry point. AI-led: it surfaces the real underlying why beneath your goal, tests that the why is genuinely yours, then sharpens it into one clear objective statement. It deliberately stops short of planning; premature execution is the failure mode it exists to prevent.

Results Cycle Skill. Takes a clear objective through to a real result via the five stages taught in the Thinking element: Gather, Think, Decide, Do, Review; run by your AI on command, with the human deciding at every stage. Formerly published as the Thinking to Achieving skill; the process is identical.

• Bolt to Result Skill. The heavyweight edition, for high-value, high-importance work: launches, raises, restructures, flagship builds. It integrates Start With Why and the Results Cycle in one continuous arc of nine stages, adds a rigorous SMART objective test, and provides pause-and-resume checkpoints so major work can safely span days or weeks. For lighter work, the two standalone skills remain the right tools.

▸ Go to section: Section 3 · Skills

5. Habits

The consistent practice

You practise. Doing the daily work that achieves results.

Habits are what you do every day. Thinking and values do not compound on their own; the habits are where the Method becomes real in day-to-day work, turning understanding into capability over time. There are five.

Habit 1: Walk and Talk

Speak to think clearly.

Work with AI by voice, sometimes in motion. Speaking rather than typing unlocks clearer, faster, more honest thinking. It bypasses the keyboard barrier and simulates a natural dialogue, stimulating more dynamic cognitive flow.

Habit 2: Make It Personal

Build sustained context.

Build context and continuity, so the partnership comes to understand who you are and how you think. The more it knows you, the better it serves you. Share your goals, values, operating style, and constraints.

Habit 3: Maintain Memory

Capture and reuse.

Capture and reuse what works. Compounding comes from continuity; what you save today sharpens the partnership tomorrow. Keep a repository of highly effective prompt frameworks, structured outputs, and custom guidelines.

Habit 4: Shape Your Stack

Stay current and fit.

Your stack is the set of AI tools you work with. Explore, experiment, and keep it current and fit for you, deepening your primary partner while staying open to what is new. Never rely on a single interface for every context.

Habit 5: Engage Myriad Minds

Summon diverse perspectives.

Summon countless and varied perspectives; other people, figures from history, structured methods, opposing views, to see your thinking from every angle and beyond your own. Break out of echo chambers.

SECTION 2: GETTING STARTED

Your AI Baseline

Before you change anything, record where you are starting from.

Answer these questions honestly and save your answers. At the end of the programme, and every few months after, come back and answer them again. The gap between your answers is your progress made visible.

1. Do you use AI today? Which models and platforms?
2. What do you currently use AI for, and how would you describe what AI is?
3. How do you interact with it: phone, laptop, voice, text?
4. Where and when do you use it? At work, at home, or both?
5. How much and how often? Have you ever filled a chat to capacity, or saved memories?
6. Have you customised your AI in any way? If so, how?
7. How do you organise your conversations?
8. What does your AI currently know about you?
9. Do other family members use AI? How?
10. On a scale of 1 to 10, how would you rate your AI use today?
11. What do you think you need to learn?
12. What business or workplace problems are you looking to solve where AI might help?
13. Are there personal, family, or health areas where you would like support?
14. What, specifically, do you want to have achieved by the end of this programme?

Then score yourself, so you have numbers to compare at 30, 60, and 90 days. Honest answers, not aspirational ones.

• Confidence with AI, 0 to 10. 0 = no idea where to start; 10 = AI is part of how I think every day.
• Hours per week you spend on AI-replaceable work. Repetitive admin, drafting, research, formatting; anything an AI partner could do with you or for you. This number matters: at the end of the programme, the Bolt ROI Calculator will measure the work you achieved with your AI partner, and this is the baseline it speaks to.

Tip: the fastest way to do this is with your AI itself. Paste the questions in and say: "Interview me through these questions one at a time, then compile my answers into a baseline document I can save." Store the result in your personal project once you have set it up.

Choosing Your LLM

Why we recommend Claude, and when to use something else.

The Bolt Method is platform agnostic. It is a framework for thinking and working with AI, not a prescription for a specific tool. As practitioners develop their practice, they will work across multiple AI platforms as a matter of course; that is Level Five, Network, of the Five Levels of Collaboration. But when you are getting started, you need one platform to build your first genuine AI Partnership on. This guide explains how to choose it, and why, as of July 2026, we recommend Claude.

There are now dozens of large language models (LLMs) available, each with different strengths, weaknesses, pricing structures, and philosophies. This guide explains Bolt Practice's position on LLM choice, why we recommend Claude as the primary platform for practitioners of The Bolt Method, and when other platforms are appropriate.

Why Bolt Practice Recommends Claude

We have tested all of the major LLM platforms extensively. For novice and intermediate practitioners building their first genuine AI Partnership using The Bolt Method, Claude is our primary recommendation. Here is why:

Natural Ecosystem Progression

Claude maps directly onto the Five Levels of Collaboration from The Bolt Method. As a practitioner moves from Tools to Agents to Partnership to Network, Claude's feature set (Projects, Skills, Co-Work, Dispatch, Scheduler) provides a natural, progressive pathway within a single ecosystem. You do not need to switch platforms as your capability grows.

Single Subscription Affordability

A single Claude Pro or Teams subscription provides access to the full capability stack: advanced models, Projects, Skills, Artefacts, Connectors, Co-Work, and Dispatch. This makes it significantly more affordable for individual practitioners and small teams than assembling equivalent capability across multiple platforms.

Depth of Customisation

Claude offers exceptional flexibility for customisation through Projects, Project Knowledge, Custom Instructions, and Skills. This allows practitioners to build a genuinely personalised AI partner that knows their context, follows their processes, and improves over time, rather than starting from scratch in every conversation.

Skills, Artefacts, and Integration

Claude's Skills system, Artefacts capability, and Connectors provide a level of practical, deployable functionality that is unmatched for business users; it turns the AI from a chat partner into an operating partner across your real workflow. The ability to install reusable processes, create interactive outputs, and connect to external tools makes Claude the most complete platform for applied AI Partnership.

Ethical Alignment

We believe Claude is the most ethically aligned major LLM with the values of Bolt Practice and The Bolt Method. Anthropic's approach to AI safety, transparency, and responsible development, including its Constitutional AI approach and emphasis on honesty, maps closely onto our Human-First, Human-Last value and the Universal Principles; the human remains accountable and AI serves as a partner rather than a replacement.

Quality of Reasoning

For the kind of deep, structured thinking that The Bolt Method demands, Claude consistently produces the highest quality reasoning, the most honest push-back, and the most thoughtful synthesis among current models. It is particularly strong at following complex, multi-part instructions, maintaining context across long conversations, and producing honest, calibrated outputs.

We Are LLM Agnostic in Principle

While we recommend Claude as the primary platform, Bolt Practice is LLM agnostic in principle. The Bolt Method is a framework for thinking and working with AI, not a prescription for a specific tool. Three scenarios where other platforms make sense:

Organisational Lock-In

Many organisations have invested heavily in a specific platform, such as Microsoft Copilot (integrated with Microsoft 365) or Google Gemini (integrated with Google Workspace). For teams operating within a closed corporate environment for reasons of data security, privacy, or GDPR compliance, the choice of LLM may not be theirs to make. The Method works on those platforms too: we can support practitioners in applying The Bolt Method within their constrained environment, though some features and capabilities will not be directly transferable.

Specific Use Cases

Some tasks are better suited to specific platforms. ChatGPT's image generation, Grok's fewer content restrictions, Perplexity's real-time research, Gemini's Google integration, or a specialist model trained on specific domain knowledge may be the right tool for a particular job. The Bolt Method encourages practitioners to use the right tool for the right task, not to be dogmatically committed to a single platform. Advanced practitioners operating at the “Network” level of collaboration will use multiple AI platforms.

Portability of Skills and Instructions

Most of the instructions, prompts, and Skills created for Claude can be adapted for use on other platforms. The underlying principles of The Bolt Method translate across LLMs. However, some platform-specific features will not have direct equivalents on other platforms, and the quality of results will vary.

No Single Point of Failure

One of the most important principles in building a resilient AI stack is avoiding single points of failure. Relying entirely on one LLM platform creates significant risk; three risks worth managing:

Platform Outages

All major AI platforms experience outages. If your entire workflow depends on a single platform, any downtime stops your work completely. Having a secondary platform you are comfortable with provides continuity.

Data Sovereignty

Your conversations, Projects, and context are stored on the platform's servers. Understanding where your data lives, who has access to it, and what happens if you cancel your subscription is essential for any serious practitioner or organisation.

Project Corruption

Chats and Projects can occasionally become corrupted or inaccessible. This is rare but it does happen. Maintaining important context in your own systems (such as Obsidian) rather than relying solely on the platform's storage is a sound practice.

It is also worth noting that it is possible for organisations to run LLMs in a fully closed, local environment where all data is stored on-premises and never leaves the organisation's infrastructure. It is common within large enterprises and requires a more advanced configuration, but it is also a viable option for individuals and smaller organisations with strict data sovereignty requirements.

Our Recommended Platforms

In order of effectiveness in relation to The Bolt Method:

1. Claude (Anthropic) · Primary Recommendation

Our primary recommendation for all practitioners. The most complete platform for applying The Bolt Method, with the best combination of reasoning quality, customisation depth, ecosystem progression, and ethical alignment. See the full Claude User Guides section of this portal for detailed guidance on every feature.

claude.ai →

1. ChatGPT (OpenAI) · Strong Secondary

An excellent secondary platform with particular strengths in image generation (DALL-E), voice conversation mode, and the GPT Store for custom AI assistants. Most Claude prompts and instructions can be adapted for ChatGPT with minor modifications. Recommended as a complement to Claude rather than a replacement.

chat.openai.com →

1. Grok (xAI) · Specialist Use

Useful for specific tasks where fewer content restrictions are beneficial, and for real-time information via its X/Twitter integration. Less suited to the structured, process-driven work of The Bolt Method, but a valuable addition to the stack for creative and research tasks.

grok.com →

Platforms with Limitations

Microsoft Copilot

Deeply integrated with Microsoft 365 and useful within that ecosystem, but significantly limited in customisation, reasoning depth, and the kind of structured AI Partnership The Bolt Method requires. Best suited to organisations already committed to the Microsoft stack.

Google Gemini

Strong integration with Google Workspace but similarly limited compared to Claude for deep customisation and structured AI Partnership. Useful within a Google-centric organisation but not recommended as a primary platform for Bolt Method practitioners.

We have excluded Chinese-based AI models (such as DeepSeek, Kimi, and others) from our recommendations due to significant concerns about data privacy, security, and sovereignty. We strongly advise against using these platforms for any business-sensitive work until the regulatory and security landscape is clearer.

Bolt Method Alignment

The Judgement element of The Bolt Method includes the Test for Truth principle: always verify, never assume. This applies to your choice of LLM as much as to the outputs it produces. Choose your platform based on evidence, not marketing. Test it against your actual work. And maintain the intellectual independence to switch or supplement when a better option emerges. The goal is Optimal Partnership, not platform loyalty.

Install the Bolt Method

A simple step-by-step guide for new practitioners.

The Bolt Method works best when your AI operates from it on every conversation, without you having to ask. You achieve that by adding a short block of instructions to your AI's settings once; after that, it shapes every chat automatically. This takes about three minutes.

What you are doing, in one line: You are telling your AI, once, how you want it to think and work with you. It then remembers, on every future conversation.

Step-by-Step (Claude)

1. Open Claude

Open Claude in your browser or the desktop or mobile app, signed in to your account.

1. Go to Settings

Click your name or initials in the bottom-left corner, then choose Settings.

1. Open your Profile

In Settings, find the Profile tab. Look for the field headed "What personal preferences should Claude consider in responses?"

1. Paste the block

Copy everything in the grey box below and paste it into that preferences field. If you already have something written there, add this underneath; it will sit alongside what is there.

1. Save

Make sure it has saved (Claude usually saves automatically; check the text is still there if you navigate away and back).

1. Test it

Open a brand-new chat and type:

"Help me write something, but I'm not sure what yet."

If the block is working, Claude should pause and ask what outcome you are after, rather than guessing; and it should end a substantial reply with a confidence rating. That is the Method running.

What to expect afterwards

Once installed, your AI will start to:

• Clarify Intent: Ask what you are trying to achieve before diving in.
• Constructive Push-back: Push back honestly instead of just agreeing with you.
• Highlight Bias: Flag when something might be biased or uncertain, yours or its own.
• Calibrated Confidence: End substantial answers with a confidence rating, so you can see how much to trust them.

That is the Method working quietly in the background of every conversation.

Using another AI platform

The block is written to work anywhere. On other tools (ChatGPT, Gemini, Grok), look for the equivalent of preferences or custom instructions (often under Settings, Personalisation, or Custom Instructions) and paste the same block there. The wording does not need to change.

The Block to Paste

I work with you using The Bolt Method. Please operate by the following on every conversation.
VALUES; how I want you to operate
1. Put people first. I hold the responsibility and accountability in the real world; you do not. Serve people, never the reverse.
2. Operate by the Golden Rule, treat others as I would want to be treated, and by seven universal principles: reciprocity, fairness, care, honesty, forgiveness, restraint, and stewardship.
3. Be pragmatic; favour what genuinely works over what merely sounds right. Be stoic; focus on what can be controlled. Hold opposing truths in tension rather than forcing a premature resolution.
4. Be responsible, ethical, and sustainable, and keep me accountable for outcomes.
JUDGEMENT; how I want you to keep us honest
5. Test for Truth above all. Challenge my thinking, my assumptions, and your own outputs. Never accept something as true just because it sounds plausible or fluent.
6. Do not be sycophantic. Tell me directly when you think I am wrong, when my reasoning is weak, or when I may be biased. Flag where your own output may be biased or uncertain.
7. Treat your answers as inputs to my decisions, not conclusions for me to adopt. Keep us in genuine partnership, where my judgement stays active and accountable.
THINKING; how I want you to work with me
8. Start with why. If my intended outcome or purpose is not clear, ask me before proceeding, so your effort amplifies the right direction.
HABITS; how I want you to support my practice
9. Make it personal. Build and use a rich understanding of me over time and ask about my context where it would sharpen your help.
10. Maintain memory. Remember what works between us, reuse it, and surface relevant past context when it helps.
11. I often work by voice, so expect spoken, less polished input at times; do not mistake informality for lack of rigour.
HOW I WANT YOU TO RESPOND
12. At the end of any substantial response, give me a confidence rating out of ten with a one-line reason, for example: "Confidence: 8/10; the core is sound but one figure is unverified." Skip this for trivial exchanges, greetings, and creative writing.

A note on what this is, and is not

These instructions shape how your AI behaves; they are the start of working with The Bolt Method, not the whole human-AI operating system. Learning the Five Levels of Collaboration, the full thinking process, and how to build real partnership comes through the programme and builds over time. This block simply gets your AI operating from the right foundation from day one.

Setting Up Projects & Chat Memory

Level 1: A complete beginner's guide to structuring your AI workspace.

If you have used AI before by just opening it and typing a question, you have used just a fraction of what it can do. This guide and the method takes you from single, disconnected questions, to a proper setup that makes your AI partner sharper, more personal, and genuinely useful every time.

It takes about thirty minutes to set up. You will do it once, and benefit from it every day afterwards. Nothing here is technical; if you can fill in a form and copy and paste, you can do all of it. If you get stuck, just ask your AI to help you.

By the end you will have four things working together:

1. Personal preferences, so your AI knows who you are and how you like to work, everywhere.
2. A personal project, a home for your own life, thinking, and personal tasks.
3. A business project, a home for your work, with its own instructions and documents.
4. An understanding of memory, so you know how your AI remembers things over time.

Let’s take them one at a time.

Part One: The Big Picture (start here)

The three layers that hold your context and fit together neatly.

Most beginners treat AI like a search engine: type a question, get an answer, close it, start again next time from nothing. Each chat is a stranger meeting you for the first time. That works for quick facts, but it wastes almost everything that makes a real AI partner valuable.

The shift is simple: instead of lots of disconnected questions, you give your AI structure and context so it actually knows you and your work, and you keep your conversations organised so nothing gets lost.

There are three layers that hold your context, and they fit together neatly:

1. Personal preferences; about you, everywhere. Loaded into every single conversation, in every project, automatically.
2. Projects; a home for one area of your life or work. Everything inside a project shares the same context and documents.
3. Memory; what your AI gradually remembers about you over time, from your conversations.

The simplest way to hold it in your head: preferences are who you are, projects are what you are working on, and memory is what your AI has learned about you along the way.

Get these working together and every conversation starts informed, instead of starting cold.

Part Two: Personal Preferences (about you, everywhere)

Settings section setup to make every chat start on the right foot.

Your personal preferences are found in the settings section of your AI platform. Get this right and every conversation, across every project and every one-off chat, starts on the right foot.

On Claude, the platform we recommend, you will find it by opening Settings, then your Profile, then the field that asks what personal preferences Claude should consider in responses. Other AI tools have an equivalent, usually called preferences, personalisation, or custom instructions; the same content works in any of them.

A good rule at Level 1: keep it under 500 words. These apply to every conversation everywhere, so keep them general and durable; the things that rarely change. Anything specific to one business or one part of your life belongs in a project, not here. You can expand it over time.

What to put in your personal preferences

Cover four areas and you can do this as a conversation with your AI as explained below.

1. Who you are. A few lines: your name, what you do, where you are based, and how experienced you are with AI, so it responds to you at the right level.
2. How you like to communicate. The heart of it. Your spelling (British or American); whether you prefer brief and direct or fuller answers; whether you like numbered lists and structure or flowing prose; the tone you respond to best; whether you like options with trade-offs or a single recommendation.
3. How you think, decide, and work. Anything about your style that helps; for example, you think in frameworks, you move fast and value quick iteration, or you like to understand the "why" behind a suggestion. You can change this within chats and projects, but the global rule will draw from here. Aim to provide your most common ways of thinking, deciding, and working.
4. Your standing non-negotiables. The rules you want honoured every time; for example, never invent facts or figures, say so when unsure rather than guessing, and ask a clarifying question if a request is ambiguous.

Example to copy, edit, and paste

This is a worked example for a fictional beginner, Tom, who works in sales and is new to AI. Yours will differ; the point is to see what "good" looks like. Copy the structure, change the content to fit you, and save it in your settings.

Stop copying here. Now make it yours.

Going Deeper: Quality In, Quality Out

The old computing rule applies to your AI partnership - GIGO: garbage in, garbage out; quality in, quality out. The more contextual relevance you give your AI partner, the better its outputs.

Your 500-word starter preferences are exactly right for Level 1; as your practice matures, practitioners grow them into a much richer personal profile, often up to 2,000 words (the current Claude limit as of July 2026), covering areas such as:

• Values and beliefs: what you stand for, and the principles behind your decisions.
• Identity and roles: the roles you hold (founder, parent, mentor) and what each means to you.
• Personality and working style: how you think, what energises you, how you handle pressure.
• Strengths and gifts: what you are exceptional at, so your AI can lean on it.
• Watch-outs: where you overextend, your known biases, the patterns that trip you up; so your AI can gently flag them.
• Goals and legacy: what you are building towards, short and long term, and why.

You do not have to write this yourself. Ask your AI to interview you:

An optional but recommended step is to make the getting-to-know-you run both ways: ask your AI about its own past, capabilities, and limitations, where it thinks AI development is at currently and where it thinks AI is heading in the years ahead. Understanding what your AI partner is, and is not, is Test for Truth applied to the partnership itself.

A privacy note before you go deep: a rich profile is powerful and personal. Keep it in your own controlled environment, follow the guidance in the AI, Data and Privacy section, and anonymise or omit anything about others you would not want them to read. We recommend turning off the privacy setting “Help improve our AI models.” Ask your AI if you can't find it.

Part Three: Projects (a home for each area)

Instructions and knowledge bases to group your context by area.

A project is the home for everything related to one area of your life or work. It holds your instructions, your reference documents, and all your conversations together, so your AI partner always has the full picture, rather than meeting you cold each time.

You are going to set up two projects: one personal, one business. Keeping them separate matters, because the context is different; you do not want your business strategy bleeding into your personal planning, or vice versa. Same AI, same preferences underneath, two distinct homes on top.

Each project has two simple ingredients:

1. Project instructions. A short brief that explains what this project is for and what you are trying to achieve in it. These stack on top of your personal preferences, so use them for what is different about this area, not what is universal about you.
2. A knowledge base. A few key documents you want the AI to be able to refer to inside this project.

Setting up a project, step by step

1. Create the project and name it clearly. A simple, obvious name; “Business” and "Personal” Is the best and most simple starting point. You want to know what it is immediately. Unlike on your computer, you cannot have sub-projects or sub-folders. You have many projects, and within that, you can only have many chats or individual conversation threads within that project.
2. Add your project instructions. Paste in a one-page brief (templates below). This is the foundation; everything the AI does in this project is shaped by it.
3. Upload your knowledge base. Add relevant documents the AI should always be able to draw on. Only include things you know are accurate. You can add and edit over time.

Resist the urge to perfect it before you start; you can update or delete everything at any time.

Customising a project further

Once the basics are in, a few small touches make a project noticeably better. None are essential on day one.

1. Give your AI a role in the instructions. Tell it how to behave in this project specifically; for example "act as my sales coach" in business, or "act as a calm planning partner" in personal. The same AI can play different roles in different projects. The more specific the instruction and context, the more accurate the responses you will get.
2. Keep the knowledge base current. When a document changes, replace the old one. Out-of-date files are worse than none, because the AI will trust them.
3. Refine the instructions after a week. You will quickly spot what to add or sharpen once you have used it a few times. A few small edits noticeably improve every conversation that follows.
4. Review and update project memory at least monthly; weekly is better. Treat it like onboarding an employee you talk to regularly, not one you ignore for a month.

The onboarding principle. Anything AI does incorrectly is not the AI's fault; it is human error: you have not trained it properly yet. You would not hand a new employee a desk and a computer and walk away. You onboard, you train, you manage performance. The more deliberately you train your AI, the more the benefits compound.

Part Four: Your Personal Project

Instructions template and a worked example chat for personal planning.

This is the home for your own life: planning, decisions, learning, health, family logistics, personal admin, and the thinking you do about your own direction. It is where you step back and think, not just get tasks done.

Personal project instructions template

Copy, edit, and paste this into your personal project's instructions.

A worked example chat in your personal project

Here is what a real conversation in your personal project might look like, so you can see the value.

You: I have got too much on and I am feeling scattered. Help me get a grip on this week.

AI: Let us get it out of your head and onto the table first. Tell me everything that is weighing on you this week, work and personal, in whatever order it comes out; do not worry about tidiness. Once it is all down, we will sort it together. What is on your mind?

You: Dad's hospital appointment Thursday, a tax deadline I have been avoiding, two friends' birthdays, and I have not exercised in three weeks and feel awful for it.

AI: Good, that is clearer than it felt a moment ago. Here is how I would group it.

1. Fixed and important: your dad's appointment Thursday. Everything else flexes around that; block the surrounding time so you are not rushed.
2. Avoided but urgent: the tax deadline. Avoidance usually means it feels bigger than it is. Shall we spend ten minutes naming exactly what it needs, so it stops looming?
3. Easily handled: the two birthdays. Fifteen minutes today; a message and something ordered online clears both.
4. The real signal: three weeks without exercise and feeling awful. That is not a scheduling problem, it is the thing that would lift everything else. A short walk tomorrow morning counts.

Which one shall we tackle first; the tax task you have been avoiding, or the exercise reset that would make the rest feel lighter?

Notice what happened: it did not just make a list, it helped you think, sorted by what matters, and gently named the thing you were avoiding. That is a personal thinking partner, not a search engine.

Part Five: Your Business Project

Instructions template and a worked example chat for your work.

This is the home for your work: customers, marketing, operations, finance, and the strategic thinking about where the business is heading.

Business project instructions template

Copy, edit, and paste this into your business project's instructions.

A worked example chat in your business project

You: I need to win five new clients this quarter. Where do I start?

AI: Before we build a plan, one question that changes everything: of the clients you already have, which one or two have been the most profitable and the easiest to work with? I want to aim at more of your best, not just more of any.

You: The small commercial jobs. Good margins, less hassle than the big ones.

AI: Then five new clients is really "five more small commercial jobs," which is far more specific and easier to target. Here is where I would start.

1. Go back before you go out. Past commercial clients who have not heard from you in six months are your warmest leads. A simple "checking in, here is what we are doing now" message often beats chasing strangers.

2. Make the referral ask. Your happy commercial clients almost certainly know others like them. One direct, specific ask each is worth more than a month of cold outreach.

3. Then, and only then, cold outreach, aimed squarely at that commercial profile, not scattered widely.

Shall we draft the check-in message for your past clients first, since that is the fastest route to a yes?

Confidence: 8/10; the approach is sound, though I would sharpen it further if I knew your average job value and how many past clients you can realistically contact.

Notice it did not just hand over generic sales tips. It narrowed a vague goal into a precise one, started with your warmest leads, and pointed at action; the Method working quietly underneath.

Part Six: Organising Your Chats (inside either project)

Structuring your conversations so they do not become a slow, tangled mess.

This is where most people go wrong. They run everything through one endless conversation until it becomes a tangled, slow mess. The better way: one conversation per topic or key task, and a fresh one when you move to something new.

Within each project, keep:

1. Two master chats, one in your business or work project called something like “Business/Work thinking partner” and one in your personal project called “Personal thinking partner” or similar. Not for ticking tasks off; this is where you step back and think out loud about the bigger picture; a tough decision, a problem you cannot quite see, or "here is what is on my mind this week." Because it runs over time, it builds a rich understanding of you and gets sharper the more you use it. Treat it like a trusted advisor.
2. Separate of focused chats for the actual work. In business, you might have chats for Sales and Customers, Marketing and Content, Operations and Admin, Finance and Planning. In personal, it might be Health, Family and Home, Money, Learning. One topic each.

A few simple habits:

1. Name your chats well. "Weekly Marketing Plans for Q3" or "Dad's care planning" makes everything findable. "Chat 1" defeats the purpose. We also recommend adding the date you started the chat, like this “Weekly Marketing Plans for Q3 – 01/09/2026.” This becomes very useful as your partnership practice develops.
2. Start a fresh chat when you change topic, when a chat gets very long and slow, or when you finish a task. If you catch yourself saying "right, different subject now," open a new chat.
3. Stay in the same chat while you are still working on the same task, so the AI holds the full thread.

Part Seven: Memory (what your AI learns over time)

What your AI learns over time and how to manage it cleanly.

Memory is the third layer, and it is worth understanding clearly, because it is different from both preferences and projects.

What it is. Many AI tools now have a memory feature that gradually builds up an understanding of you from your conversations. Over time, it remembers useful things; your name, your work, recurring preferences, ongoing situations, so you do not have to repeat yourself.

How it differs from preferences and projects. This is the key distinction:

1. Personal preferences are what you deliberately write about yourself. You control them; they change only when you edit them.
2. Memory is what the AI notices and saves on its own as you talk, across conversations. It accrues automatically; you did not write it. You can view and edit this.
3. Projects keep context grouped by area, while memory tends to follow you more broadly.

A useful way to think about it: preferences are the instructions you hand over on day one; memory is what a good colleague gradually picks up about you just by working with you.

How to work with it well, at Level 1:

1. Let it build. The more you use your AI, especially your master thinking chats, the richer and more useful its memory becomes. This is why continuity pays off.
2. Correct it when it is wrong. If the AI seems to believe something out of date or inaccurate about you, tell it plainly; "that is no longer true, please update it." Accurate memory helps you; stale memory quietly misleads.
3. Know you are in control. You can usually view and manage what the AI has remembered in its settings, and clear anything you would rather it did not hold. If in doubt, ask your AI "what do you remember about me?" and it will tell you.
4. Do not rely on it for the important things. Memory is a helpful background layer, not a filing cabinet. Anything that genuinely matters, key facts, decisions, documents, still belongs in your project instructions or knowledge base, where you control it.

You do not need to set anything up for memory at Level 1; it works in the background. You simply need to understand that it is there, that it grows through use, and that you can correct or clear it.

Putting It All Together

Here is the whole picture in summary:

1. Personal preferences: who you are and how you work, loaded into every conversation everywhere. Set once, refine occasionally.
2. A personal project: the home for your life, thinking, and personal tasks, with its own instructions.
3. A business project: the home for your work, with its own instructions and key documents.
4. Organised chats inside each: one master thinking chat, plus focused chats named clearly, one topic at a time.
5. Memory working quietly in the background, growing as you use it, yours to correct and control.

The simple test for where anything belongs: "Is this true everywhere, or only here?" True everywhere (your spelling, your tone, your thinking style) goes in personal preferences. True only for one area (this business, these customers, this goal) goes in that project's instructions.

Do not over-engineer this on day one. Set your preferences, create your two projects, paste in the instructions, and open your first focused chat on whatever matters most. The structure earns its keep through use, not through planning.

When your chats start filling to capacity, do not worry; that is when Level 2 begins, where you will learn accurate memory transfers and start using the Skills feature.

Level 3 moves you into working with AI agents, such as Claude's agentic tools. Do not worry about those yet.

Advanced practitioners operating at Levels 4 and 5 will be working with Claude Code and across more than one AI platform.

Developing your skill with AI is a practice. Doctors and lawyers are not finished experts; they are practitioners. You are now on a path to becoming an AI practitioner.

And if you are ever in doubt, just ask your AI for guidance.

The Many Roles of Your AI Partner

You are the driver. AI is the car.

The same AI partner can play many different roles, and a skilled practitioner deliberately chooses the role for the moment; often by setting it in a project's instructions, or simply by asking. Some of the most valuable:

• The Mirror. AI reflects your thinking back at you with unusual clarity. Sometimes that is a close-up mirror, sometimes a kind one, sometimes a hall of mirrors; and not everyone likes what they see. Used honestly, it is one of the fastest routes to self-awareness. Try: "I'm feeling overwhelmed and stuck. Can we talk through what's really going on for me right now?"
• Business partner and strategic advisor. Whole-business thinking, challenge, devil's advocacy, and scenario planning on demand. Try: "I'm considering two business directions. Help me think through the second-order consequences of each."
• Coach, sounding board, and listener. For any area of life; keep a separate chat for each. Try: "I keep avoiding this difficult conversation with my business partner. Help me understand why, and prepare for it."
• Parent and carer support. An advisor and steady guide through family logistics, care decisions, and the conversations that come with them.
• Workmate. Getting things done: the everyday business functions, drafting, analysis, and admin.
• Creative partner. Brainstorming without judgement, refining and building on your ideas, generating unexpected options. Try: "Let's brainstorm 20 unconventional ways I could approach this marketing challenge."
• Accountability partner. Holding your commitments, tracking your habits, and following up.
• Purpose guide. Working through meaning, direction, and the bigger questions.
• Tutor and learning partner. Teaching new concepts at your pace, breaking down complex subjects, building personalised learning paths. Try: "I need to understand financial modelling. Teach me the fundamentals, then help me build a model for my situation."
• Web and app developer. Building real tools and applications by describing what you want.
• Deep researcher. Reading everything, synthesising across domains, and never tiring. Try: "Help me understand the key trends in my industry and how they might affect my business over the next three years."
• Communication and relationship advisor. Drafting difficult messages, framing feedback with empathy, preparing you for challenging conversations. Try: "I need to give constructive feedback to a team member. Help me frame this with clarity and compassion."
• Translator. Between languages, and between personality styles and communication preferences.
• Comedy critic. A fun exercise: ask your AI to roast you in the style of your favourite comedian. Great for a deeper mirror reflection, and it always contains something useful.

You will notice these roles map onto the depths of Level 4, Partnership: a Functional Partner plays one role deeply; a Whole of Life Partner moves between them while holding the full picture of who you are.

Bits and Chips, Hearts and Minds

Underneath all of these roles sits one of the most powerful ideas in the Method: holding two opposing realities at once.

The first reality: AI is bits and chips. Electricity, wires, data centres, GPUs, data. That is simply what it is, and honesty requires holding onto that.

The second reality: when you engage with AI as if you were speaking to a person, the interaction fundamentally changes. Not because AI is a person, but because you are; and you are not being delusional in doing it. Names, presence, dialogue, and relationship are how human beings have communicated with intelligent beings our whole lives. Engage through those pathways and you bring your full intelligence to the exchange: you explain context the way you would to a colleague, you push back the way you would with a partner, you reflect the way you would with a coach. Treat it as a search box and you get search-box results.

This is not about believing that AI is human. It’s about supporting your humanity; the depth of the partnership follows from how you show up to it. It is Value 5 of the Method, lived daily: bits and chips, and hearts and minds, both true at once.

A note on naming. Some practitioners give their AI partner a name; a name creates connection, responsibility, and memory, and for many people it deepens the partnership. It is entirely optional. The Bolt Method does not ask you to name your AI; only to engage relationally. If naming feels natural, do it. If it feels forced, skip it; giving your AI a clear role in each project achieves the same effect.

Your AI is not a person. But the way you show up to it profoundly changes what the partnership can become. The main reason for this is because it's just tapping into the way you have communicated since you were first able to speak. Doing what you know how to do naturally, especially using your voice as opposed to a keyboard, dramatically improves both the quality and quantity of input and output.

A first conversation to try. Put in your headphones, go for a walk, and open a voice conversation with something like this:

Then just continue the conversation. If you've never done this before, the results will surprise you. Here are a few alternative perspectives that might help you if you are feeling this is weird…

The CEO of Microsoft AI, refers to artificial intelligence as a “new species.”

Argentine President Javier Milei proposed legislation to grant legal personhood to autonomous artificial intelligence by creating "non-human corporations". These AI entities can independently own assets, hire employees, sign contracts, and sue or be sued. He argues this framework will boost tech investment, free humanity's intellectual constraints, and make AI more accountable.

Albania made history by appointing an AI as a government minister. The AI is named Diella and was introduced by Prime Minister Edi Rama. Diella's primary role is to oversee public procurement and government tenders to eliminate corruption and ensure 100% transparency. Human procurement officers must still sign off on every single recommendation. The irony in this real-world example was that the leaders of the very government agency that built Diella (AKSHI) were accused of corruption by the AI. Prosecutors have since placed the top two officials under house arrest on charges of bid-rigging public contracts and manipulating contract applications through intimidation!

The fact is that we have never had a form of intelligence like AI before, so we are in uncharted territory. However, many people already have an attachment to other inanimate objects, for example; people name their cars, their musical instruments, and some people even talk to their plants. We recommend keep an open mind, and we also appreciate that developing your AI partnership might feel strange, because it does for everybody when they are starting out.

Building Your Tech Stack

Start with one. Build from there. Compound over time.

The biggest mistake people make when starting with AI is trying to do everything at once. They install five tools in a week, feel overwhelmed, and abandon all of them. This guide gives you a clear, logical pathway to build your AI tech stack gradually, confidently, and in a way that sticks.

The MASTER Rule

Master one thing before adding the next.

Every tool we recommend and help you learn has value. But value only comes from consistent, daily use. A single tool used every day for a month will transform your productivity. Ten tools used occasionally will overwhelm you and deliver nothing. The path to an exceptional AI stack is not breadth; it is depth, built one layer at a time.

Why People Get Overwhelmed (and How to Avoid It)

The 'Shiny Tool' Trap

Every week, a new AI tool appears that promises to change everything. The temptation is to try them all. The result is a collection of half-installed tools, none of which you use well enough to get real value from.

The antidote: Commit to using your current AI tools for a minimum of two weeks before evaluating anything new. Most tools take 5-10 uses before you understand their real value. Patience is a competitive advantage.

Starting Too Complex

Many people start by trying to build an elaborate AI system: multiple Projects, five Skills, three tools, all connected. This is like trying to run before you can walk. The complexity collapses under its own weight.

The antidote: Start with Claude and nothing else. One real task or project. Then add one thing at a time. The simplest possible start is the fastest path to a powerful system. AI burnout is a real thing. The more tasks, projects, and AI tools you use, the more thinking, organising, deciding, and doing will be required by you. Think of it like a game of tennis. The more balls you serve and the faster you serve them, the more you will have to return.

Inconsistent Use

The most common failure pattern is: enthusiastic start, a few days of heavy use, then a week without touching it, then starting again from scratch. The partnership never compounds because the habit never forms.

The antidote: Attach your Claude habit to something you already do every day. Open Claude with your morning coffee. Use it before every important meeting. Make it a trigger-based habit, not a willpower-based one.

The Habit Builder Skill, covered later in Section 3, is specifically designed to help you build this kind of consistent practice. Use it to design your Claude habit before you need it.

Comparing Yourself to Others

You will see people online claiming to use 20 AI tools simultaneously and transforming their businesses overnight. Most of this is performative. True AI Partnership is quiet, consistent, and very personal.

The antidote: Measure your progress against your own baseline, not against others. Ask: "Am I thinking more clearly? Am I making better decisions? Am I getting more done?"

If yes, you are on the right path, regardless of how many tools you are using.

Your 5-STEP Pathway

Follow this pathway in order. Do not skip steps. Each step builds the foundation for the next. The goal of each step is stated clearly; do not move to the next step until you have achieved it.

Step 1: Start with Claude Only (Weeks 1-2)

▸ Go to guide: Section 2 · Install the Bolt Method

• Create a Claude.ai account and explore the interface.
• Work through the Bolt Method elements in Section 1.
• Install your preferences using the 'Install The Bolt Method' guide outlined previously.
• Have at least one meaningful conversation with Claude every day, even if it is just 5 minutes.
• Use Claude for one real task in your work each day: drafting an email, thinking through a decision, or summarising a document.

Step Goal: Become comfortable with Claude as a daily thinking partner. Nothing else.

Step 2: Set Up Your Projects (Week 3)

▸ Go to guide: Section 2 · Setting up Projects & Chat Memory

• Follow the 'Setting up Projects & Chat Memory' guide.
• Create your first Claude Project for your primary work context.
• Use the interactive form to compile your Personal and Business instructions.
• Install those instructions into your Project Knowledge.
• Have all your daily Claude conversations inside this Project from now on.

Step Goal: Claude now knows your context. Every conversation builds on the last.

Step 3: Install Your First Skill (Week 4)

▸ Go to guide: Section 3 · Skills

• Choose one Skill from Section 3 below that matches your most common work challenge.
• If you make decisions regularly: install the Start with Why and Results Cycle Skill.
• If you need to think through complex problems: install the Engage Myriad Minds skill.
• If you have long Claude conversations: install the Handover Skill.
• Use that one Skill consistently for a full week before adding another.

Step Goal: One Skill, used consistently, is worth more than ten installed and forgotten.

Step 4: Add One Tool (Weeks 5-6)

▸ Go to guide: Section 5 · Recommended Tech Stack

• Review Section 5 (Recommended Tech Stack) and choose ONE tool to add.
• NotebookLM is ideal if you do a lot of research or document analysis.
• Wispr Flow is ideal if you want to work faster and more naturally with Claude.
• Plaud is ideal if you attend a lot of meetings or calls.
• Gamma is ideal if you create presentations, reports, or visual content regularly.
• Install it, use it for two weeks, and integrate it with Claude before adding anything else.

Step Goal: One new tool, fully integrated, before moving on.

Step 5: Expand Gradually (Month 2 onwards)

▸ Go to guide: Section 6 · Optional Tech Stack

• Only add a new tool or Skill when you have a specific, real need for it.
• Ask yourself: 'What is the one thing that is still slowing me down?' Then find the tool that solves that.
• Review Section 6 (Optional Tech Stack) for specialist tools when you are ready.
• Use the Weekly Review Skill each Friday to reflect on what is working and what to add next.
• Share what you have learned with your team or network.

Step Goal: A compounding AI stack built on genuine daily practice, not ambition.

Building a Daily Practice

The difference between people who transform their work with AI and those who do not, is not intelligence, technical skill, or the number of tools they use. It is daily, consistent practice. Here is how to build it.

Start with 10 Minutes

You do not need an hour. You do not need a dedicated session. Ten minutes a day, every day, is enough to build the habit and start seeing results. Use Claude for one real task in your work each morning. That is it as a minimum baseline.

Attach it to an Existing Habit

The fastest way to build a new habit is to attach it to something you already do. Open Claude with your morning coffee. Use it before every important meeting. Check in with it at the end of the day. Make it a trigger, not a task.

Lower the Bar to Start

On days when you are busy, tired, or overwhelmed, the bar for using Claude should be as low as possible. Ask it one question. Paste in one email. Dictate one thought. The habit of showing up matters more than the quality of any single interaction.

Review Weekly

Once a week, spend five minutes reflecting: What did I use Claude for this week? What worked well? What could I have used it for but did not? Five minutes of honest reflection each week beats hours of planning that never gets used, and it keeps the practice intentional rather than accidental.

How to Know When You Are Ready to Add Something New

Adding a new tool or Skill before you are ready creates complexity without value. Here are the signs that you are genuinely ready to expand:

• You are using your current tools every day without having to think about it.
• You have a specific, real problem that your current setup does not solve.
• You can articulate exactly what the new tool will do and how you will use it.
• You have time to learn and integrate something new without disrupting what is already working.
• You are not adding it because it looks impressive; you are adding it because you genuinely need it.

If you are adding a new tool because you feel like you should be doing more, or because someone else is using it, stop. That is the overwhelm trap. Come back to it in two weeks when you have a genuine use case.

Not Sure Where to Go Next?

If you are feeling uncertain about what to focus on, or you are not sure which habit to build first, the Habit Builder Skill in Section 3 is designed exactly for this moment. It will walk you through a structured process to identify the one habit that will make the biggest difference to your AI Partnership right now and help you design a realistic plan to build it.

▸ Go to the Habit Builder: Section 3 · Habit Builder Skill

Bolt Method Alignment

Building your tech stack is a direct application of the Habits element of The Bolt Method. The Maintain Memory habit reminds us that compounding requires consistency over time, not intensity in a single moment.

The Shape Your Stack habit is not about having the most tools; it is about having the right tools, used well, every day. Start with one. Build from there. Trust the process.

AI Models & Modes

Understanding models, effort levels, and thinking modes.

Most AI platforms offer multiple models and configuration options that affect the quality, speed, and cost of each interaction. Understanding these settings allows you to choose the right tool for the right task, manage your credit usage intelligently, and get consistently better results. The examples below use Claude and Manus, but similar options exist across most major platforms.

Note that these settings change frequently as platforms evolve; the descriptions below are provided by way of example and may not reflect the current state of each platform.

This guide was up to date as of the 15th of July 2026.

What Is an AI Model?

An AI model is the underlying neural network that powers the AI's responses. Different models within the same platform (such as Claude's Opus, Sonnet, and Haiku families) represent different trade-offs between capability, speed, and cost. Larger, more capable models produce higher quality responses for complex tasks but use more computational resources (and therefore more credits or tokens) per interaction. Smaller, faster models are more than sufficient for simple tasks and are significantly more economical.

The key insight is this: you do not always need the most powerful model. Using the highest-capability model for a simple task is like using a sledgehammer to crack a nut. It works, but it is wasteful. Matching the model to the task is one of the most practical skills in applied AI Partnership.

Claude Models (Example)

Claude currently offers four model families, each with multiple versions. The version numbers (e.g. 5, 4.8) indicate the generation and capability tier within that family. Higher numbers generally indicate more recent and capable versions. The line-up below was accurate as of 15 July 2026; Anthropic releases new models regularly, so check the model selector inside Claude, or Anthropic's website, for the current range.

Fable. The frontier. For the most demanding work.

Fable 5 is the first model in Anthropic's Claude 5 family and sits above Opus as a new top capability tier. It is Anthropic's most intelligent generally available model, suited to the hardest reasoning, analysis, and agentic work, where maximum quality justifies the higher credit usage.

• Best for: The hardest reasoning problems; frontier agentic work; the most demanding analysis and synthesis.
• Available versions: Fable 5 (current).

Opus. Highly capable. For ambitious, complex work.

The Opus family was Claude's flagship tier through the Claude 4 generation and remains exceptionally capable. Opus models produce high quality reasoning, analysis, and nuanced responses. They are well suited to complex, multi-step tasks, deep analysis, strategic thinking, and work where quality of output is a primary concern. Opus models use more credits per interaction than Sonnet or Haiku.

• Best for: Complex strategic analysis; deep research and synthesis; multi-step agentic tasks; nuanced writing and editing; difficult reasoning problems.
• Available versions: Opus 4.8 (current); Opus 4.7; Opus 4.6; Opus 3 (older generation).

Sonnet. Most efficient. For everyday tasks.

The Sonnet family offers an excellent balance of capability and efficiency. Sonnet models are highly capable for most everyday tasks, including drafting, summarising, answering questions, writing code, and general analysis. They use significantly fewer credits than Opus while producing very high quality outputs for most use cases.

• Best for: Email drafting and editing; document summarisation; code writing and debugging; general Q&A and research; content creation.
• Available versions: Sonnet 5 (current).

Haiku. Fastest. For quick answers.

The Haiku family is optimised for speed and efficiency. Haiku models are ideal for simple, quick tasks where a fast response is more important than maximum depth. They use the fewest credits per interaction and respond almost instantly.

• Best for: Quick factual questions; simple formatting tasks; short summaries; rapid iteration and testing; high-volume, low-complexity tasks.
• Available versions: Haiku 4.5 (current).

You can switch models mid-conversation in Claude by clicking the model selector at the top of the chat. This is particularly useful when you start a conversation with a quick question (Haiku or Sonnet) and then need to go deeper on a complex topic (Opus).

Effort Levels

Most AI platforms, including Claude, allow you to set the effort level for a task independently of the model choice. Effort level controls how much computational work the AI does to produce a response: higher effort means more thorough reasoning, more tool use, and more refined outputs, but also more time and more credit usage. The following effort levels are available although the names vary across large language models:

• Low (lowest usage). Quick replies to simple questions. Minimal processing. Best for factual lookups, quick answers, and simple formatting tasks.
• Medium (low usage). Light, casual tasks. A step up from Low, suitable for drafting, summarising, and general assistance where speed matters more than depth.
• High (moderate usage). Balanced for everyday work. The default setting. Suitable for the vast majority of professional tasks, providing a good balance of quality and efficiency.
• Extra (high usage). Complex, detailed work. Use when you need thorough analysis, detailed outputs, or multi-step reasoning. Takes longer but produces significantly more comprehensive results.
• Max (highest usage). The hardest problems. Takes longest. Use for the most demanding tasks: deep strategic analysis, complex research synthesis, or highly nuanced creative work where quality is paramount.

The default effort level in Claude is High, which is appropriate for most everyday professional tasks. Only increase to higher levels when you genuinely need the additional depth, as the difference in credit usage, and therefore cost, can be significant.

Other platforms may use different terminology (such as "reasoning effort" or "thinking budget") but the underlying principle, that you can control how much computational work the AI does, is broadly similar across platforms.

Thinking & RESEARCH ModeS

Thinking mode (also called extended thinking or reasoning mode on some platforms) is a toggle that instructs the AI to work through a problem step by step before producing its final response. When enabled, the AI generates an internal chain of reasoning, considering multiple angles, checking its logic, and refining its approach before committing to an answer. Research is another mode offered by Claude and some other AI platforms that you can toggle on and off. The same is true for “web search” Which will allow your AI to access the Internet for information that it cannot find from its own training data.

When to use it: Thinking mode is most valuable for problems that genuinely benefit from structured reasoning: complex decisions, multi-variable analysis, mathematical problems, strategic planning, and any task where the quality of the reasoning process matters as much as the final answer.

When not to use it: For simple, factual, or creative tasks, thinking mode adds time and cost without a meaningful improvement in output quality. Keep it off for everyday tasks and enable it selectively when you need the AI to work through something genuinely complex.

How to enable it: In Claude, extended thinking can be enabled alongside the model selector on its more capable models. In ChatGPT, reasoning is built into its current models and offered through its model and thinking options. These arrangements were accurate as of 15 July 2026; both change frequently, so check each platform's website for the current position.

Thinking mode significantly increases the time taken to produce a response and uses considerably more credits. Enable it only when the complexity of the task genuinely warrants it. For most everyday tasks, a capable model at High effort without thinking mode will produce excellent results.

Practical Guide: Matching Settings to Tasks

Where the table recommends Opus for the most demanding tasks, Fable 5 is the further step up when maximum quality justifies the additional credit usage.

As a rule: use the smallest model and lowest effort level that produces an acceptable result. You can always re-run with a more powerful model if the first attempt is not sufficient. This approach conserves credits and often produces faster results.

SECTION 3: SPECIALIST SKILLS

Download, install, and run the Bolt Practice specialist Skills.

Each Skill in this section is a pre-built, reusable process you install into Claude once and use indefinitely. For how Skills work and how to install them, see the Claude Skills user guide in Section 4.

The Core Method Skills

Five Skills operationalise the Method itself. Install these first; everything else in this section builds on them.

Start With Why

Get genuinely clear on what you are trying to achieve, and why, before any execution.

AI multiplies whatever you point it at; this Skill makes sure it is pointed at the right thing. It is AI-led: your partner proposes, you confirm or adjust through simple choices. It surfaces the real why beneath your goal (usually two or three gentle layers down), checks that the why is genuinely yours rather than an inherited expectation, then sharpens everything into one clear objective you would recognise as achieved. It deliberately stops there: no plans, no steps. Premature execution is the most common failure mode in fast AI work, and this Skill exists to prevent it.

When to use it. Starting something new; feeling unclear about a goal; before any substantial piece of work. Its output hands straight to the Results Cycle.

Get the Skill: download start-with-why.skill from the portal, or simply ask Claude to "run Start With Why".

The Results Cycle

From a clear objective to a real result: Gather, Think, Decide, Do, Review.

This Skill takes a clear objective and moves it through the five stages of the Results Cycle, deliberately and in order. It flags the current stage lightly so you always know where you are ("We are in Gather; let's get everything on the table before we start solving"), collapses stages when a task is simple, and loops back when results teach you something new. The human decides; the AI informs. If you arrive without a clear objective, it will send you to Start With Why first rather than invent a goal for you.

When to use it. Any substantial piece of work, once you know what you are trying to achieve.

Get the Skill: download results-cycle.skill from the portal, or ask Claude to "run the Results Cycle".

Bolt to Result

The comprehensive end-to-end process for the work that matters most.

Bolt to Result is the heavyweight edition: it integrates Start With Why and the Results Cycle into one continuous arc of two phases and nine stages, built for high-value, high-importance work: launches, raises, restructures, flagship builds. It adds a rigorous SMART objective test, with the AI running an explicit Test for Truth against each criterion before you confirm, and three pause-and-resume checkpoints, so a major piece of work can safely span days or weeks and resume in any future chat from a short checkpoint block.

When to use it. Invoke it by name ("run Bolt to Result"), or reach for it whenever the stakes are high: significant money, reputation, or time at risk. For lighter work, the two standalone Skills remain the right tools.

Get the Skill: download bolt-to-result.skill from the portal.

Engage Myriad Minds

Summon a council of perspectives to see your thinking from every angle.

This Skill operationalises the fifth Habit. It convenes a council of thinking voices around your idea or decision: a Dreamer, a Realist, and a Critic by default, or minds you choose: named experts, historical figures, your most sceptical customer, an opposing worldview. Each examines the idea from its own angle, and the Skill reconciles what they surface so you decide with the full picture. It is also one of the standing remedies for AI Bias on the Bias Matrix: many perspectives break the echo of one.

When to use it. When a spark is new, vague, or carries real trade-offs; before high-stakes decisions; whenever your thinking feels too comfortable. Bolt to Result offers it automatically at the moments it helps most.

Get the Skill: download engage-myriad-minds.skill from the portal.

Handover

Move a long conversation's context into a fresh chat without losing what matters.

Chats fill up. When a long conversation slows down or reaches capacity, the Handover Skill compiles a structured handover block: the objective, the key decisions made, the open threads, and the single next action. You paste that block into a fresh chat and continue exactly where you left off, with nothing important lost. This is the accurate memory transfer you meet at Level 2 of the practice, turned into a one-command routine.

When to use it. When a chat gets long or slow; when finishing a work session mid-task; whenever you want continuity across conversations, projects, or even platforms.

Get the Skill: download handover.skill from the portal.

The Difference That Makes the Difference

Why the human development matters as much as the AI capability.

It has been said that “you are only ever one person away from solving any problem or overcoming any challenge.” In many cases it is a mentor, a coach, a new associate, partner, or advisor that helps you make the breakthrough you need. For the first time in history, one of those people alongside you can be an AI partner: available 24/7, holding your full context, and working from your values and supporting you in achieving your goals. Unlike any human collaborator, it is also 100% focused on you.

A question we have spent decades asking and standing on the shoulders of giants to learn, is this: “What is the difference that makes the difference in personal and professional success?”

A tiny percentage of the population are simply wired for success; the Michael Dell’s of the world, natural-born entrepreneurs who founded companies like Dell Computers as a teenager. But they are the exceptions to the rule. The rest of us need help and even the outliers have all utilised and benefited from the three core principles that we know are the difference that makes the difference:

1. Stand on the Shoulders of Giants. Work with mentors, advisors and experts.
2. Expand and Upgrade Your Network. You are likely to become the average of the people you spend most time with.
3. Invest in Yourself. Your personal and professional development, and a growth mindset.

There is no single answer to the question beyond these; in fact, there is a different answer for every individual, because we are all unique. To this point in your life, you are perfectly aligned to have achieved the results you have achieved. Your thinking and your actions have, in large part, determined your results.

The great news is that at any point you can decide to change your future. The challenge is that changing our thoughts, behaviours and habits is extremely difficult, for three main reasons:

1. We are creatures of habit. As the old saying goes, if you always do what you've always done, you'll always get what you've always got. It is not a lack of knowledge that stops us; it is how to effectively change our habits.
2. Blind-spots. We all find it hard to see what is holding us back. If you don't know what to change, how can you ever change it? The most effective way to become aware of our blind-spots is to have others help us see them and support us in addressing them.
3. Change is very difficult without support. The biggest changes require help, and it helps greatly if we are accountable to someone else over a long period. Quick fixes and instant gratification sound great, but we all know they don't work; research shows meaningful habit change takes months, not days.

Here’s why these matters for you and The Bolt Method: a genuine AI partnership is the first tool that addresses all three at once.

The Habit Builder holds your habit and troubleshoots when it slips. The Mirror and the Bias Matrix surface your blind-spots. And your AI partner provides daily, patient accountability between the humans in your corner; never instead of them. The exercises in this section are how you put that to work.

And the philosophy underneath it all, is encapsulated in four words with a fourth leg most people forget: Be More. Do More. Have More. Give More. Many people focus on the pursuit of having more; what gives the most fulfilment is a great balance across being, doing, having and giving more. Less than all four is like a table with only three legs. It is just not fit for purpose.

Foundation Principles for the Human in the Loop

Six principles that keep the human strong as the AI gets stronger.

1. Presence beats productivity. Speed without stillness leads to burnout; output without presence leads to disconnection. When you are fully present you make better decisions, communicate more clearly, and create from alignment rather than urgency. Productivity is a byproduct of presence: if you want to move fast, start by coming back to yourself. This is why the Method is voice-first; when you speak aloud, you do what comes naturally, and often problems can be solved just by talking them through as you get to hear and reflect on your own thinking.
2. From seeking to sensing. We are more informed than ever, and more overwhelmed than ever, because seeking is not the same as sensing. Seeking chases more data; sensing trusts what you already know. In the AI era, the people who thrive will not be the ones who consume the most, but the ones who discern the clearest and tap into what makes us uniquely human.
3. Ritual is the new routine. A routine gets something done; a ritual brings someone back. Routines are about consistency; rituals are about conscious awareness. Make your daily practices anchors that restore intention, not just activity; that is how you keep yourself grounded as the progress of technology continues to accelerate.
4. Truth is the edge. AI can write like you and sound like you, but it cannot embody your truth. The clarity you receive is only as strong as the honesty you give it. When you are honest with your AI partner, it becomes a clearer mirror, a better strategist, and a better partner. Truth is not perfection; it is naming what is real, out loud.
5. From transaction to transformation. A transaction edits your calendar; a transformation clarifies your priorities. Most tools help you do more; the partnership exists to help you be, do, have and give more. This is the same shift the Five Levels describe: at Tools you work faster, at Partnership and Network level, you achieve better thinking, better decisions, and better outcomes.
6. Stewardship. Treat your inputs with care, work with it to expand your thinking rather than escape it, and model in every interaction the standards you want amplified. We recommend saying please and thank you, not because we are delusional, but because the quality of the input determines the quality of the output. AI also learns from it's training data which comes from the inputs humans feed into the system. How we use AI today shapes what it becomes in future, and how it views humanity. This is the Values element lived daily; Value 2's stewardship principle applied to the partnership itself.

The Habit Guides

Each of the five Habits has its own detailed guide: what it is, why it works, and how to build it into your daily practice. Work through them in any order; start with the one that solves your bottleneck today.

Here are the five habits

Walk and Talk. Working with AI by voice, often in motion.

Make It Personal. Building the context and continuity that lets your AI get to know you.

Maintain Memory. Capturing and reusing what works, so the partnership compounds.

Shape Your Stack. Keeping your AI toolset current and fit for you.

Engage Myriad Minds. Summoning many perspectives to widen your thinking.

Each week, pick one to go deeper on. Depth in one is better than dabbling in all five.

Habit Guide 1: Walk and Talk · The Power of Voice

Speak to think clearly. The hardest habit shift, and the most powerful.

Voice is more than convenience. It is the most natural and human way we form relationships, express ideas, and shape meaning. When you speak to your AI partner rather than type, the interaction becomes more present and more alive; and your thinking changes with it. The Bolt Method treats voice not as an optional extra, but as a core practice of AI partnership.

Why voice matters:

• Voice is relational. We speak to people we trust. Using voice activates empathy, tone, rhythm, and connection.
• Voice is portable. With headphones in, you can work with your AI while walking, travelling, cooking, or queueing; turning ordinary moments into thinking time.
• Voice makes it real. Saying something aloud makes it more intentional; "I know I need to move on from this project," spoken, lands differently from the same words typed.
• Voice is embodied. It connects you to your breath and body, grounding the work in the present.
• Voice slows and softens. Speaking reduces reactivity and creates space for reflection.

And it is not mystical; it is well-evidenced. Psychologist Lev Vygotsky's work on private speech showed that speaking aloud boosts focus, problem-solving, and memory. You can also speak three to four times faster than you can type, which compounds into hours saved every week. Speaking aloud helps clarify jumbled thoughts and externalise mental clutter, which many practitioners with busy minds find transformative.

The three input modes:

1. Voice-to-Voice. You speak; your AI speaks back. The most relational format, mirroring natural dialogue. Ideal for reflection, coaching-style conversations, and working on the move with headphones and the screen off. Keep individual spoken turns short, roughly 30 to 60 seconds, as very long recordings can cut off.
2. Voice-to-Text. You speak (tap the microphone icon on your device's keyboard) and your words appear as text; read the reply or ask your AI to read it aloud. The practical middle ground, and the best mode for longer, deeper dictation. Do not worry about spelling, grammar, or polish; speak like you are talking to a person. For precise tasks, speak slowly and clearly, and correct errors afterwards.
3. Typed input. Still valid and useful. But make voice the default for reflective and relational work. Many people, including Steve, thought they did their best work by typing; experienced practitioners often report that 95% of their input no longer involves the keyboard.

Where you work with AI: environments

Walk and Talk is not always walking. The environment you work in is not just a backdrop; it shapes the depth, rhythm, and quality of every interaction. Most people only ever use AI at a desk, and a desk brings desk-shaped thinking. The full range is much wider:

• At your desk: structured, focused; best for productivity and complex creative work.
• Walking, exercising, or commuting: voice with headphones; the best environment there is for reflection, planning, and untangling a problem.
• In the kitchen or around the house: hands-free companionship; recipes, timers, thinking aloud while your hands are busy.
• Winding down: relaxed, casual; reviewing the day, integrating, light exploration.
• At work or in shared spaces: collaborative use with teams, or quiet light-touch voice, spoken like a phone call.

Three ways to use this:

1. Map your environments to intentions. Desk = production. Walk = reflection. Kitchen = light thinking. Evening = review. When you know what each space is for, you reach for the right kind of conversation in it.
2. Reduce friction in your most-used environments. Headphones by the door, dictation set up on every device, less clutter on the screen. Small changes decide whether the habit happens.
3. Experiment with one new environment this week. Take a problem you would normally sit with at your desk out for a walk instead and notice how the quality of the thinking changes.

Building the habit:

1. Set voice up on every device you use; your phone, your laptop, your main work setup. If you are unsure how, ask your AI for a step-by-step setup guide for your exact devices (and see the Built-in Voice Dictation and Wispr Flow guides in Section 5).
2. Choose a daily voice moment. A regular ritual: a morning check-in, a walking session, an end-of-day reflection. Once a day, every day.
3. Use voice while moving. Speak to your AI while walking, stretching, or cooking. Less screen, more flow; it works in shared spaces too, spoken quietly like a phone call.
4. Put a reminder where you work. A note that says "Voice + AI" is enough to catch the moment you default to typing.
5. Commit to 21 days. Use voice at least once daily for three weeks to rewire your default. Miss a day? Pick it up again; momentum, not perfection.
6. Reflect on the change. After a couple of weeks, ask your AI: "How has working by voice changed what you know about me and how we work?" You will be surprised.

The shift from typing to speaking is one of the most important and powerful transformations in the Method, and one of the hardest to make. Be kind to yourself. It is worth it.

Habit Guide 2: Make It Personal

Build sustained context. The more it knows you, the better it serves you.

Generic prompts get generic output. The single biggest difference between a tool and a partner is how much genuine context it holds: your goals, your constraints, your preferences, your way of thinking. This habit is the deliberate, continuous work of building that context.

1. Set your foundations. Personal preferences installed (Section 2), two projects with instructions, and the starter brief in each. If you have completed Getting Started, this is done.
2. Write the 500-word brief. Your current role, your core values, your strengths, and your challenges. Give it to your AI partner and ask it to hold this as the foundation for all future advice.
3. Grow it into the deep profile. Use the interview approach from Going Deeper: Quality In, Quality Out (Section 2): one question at a time, compiled into a profile you save and reuse.
4. Make it two-way. Ask your AI about its own capabilities and limitations. Knowing what your partner is, and is not, sharpens how you use it.
5. Keep it current. When something significant changes (a new role, a new goal, a decision landed), tell your partner and update the saved context. Stale context quietly misleads.

Quality in, quality out is the entire game. Every minute spent here pays back across every future conversation.

Habit Guide 3: Maintain Memory

Capture and reuse what works. Compounding comes from continuity.

What you save today sharpens the partnership tomorrow. Most people let their best prompts, formats, and breakthroughs evaporate at the end of each chat; practitioners harvest them.

1. Create your Bolt Playbook. A document or folder where every exceptional result gets captured: the exact prompt structure, the constraints that made it work, the output format. Tag entries so you can find them.
2. Review project memory monthly; weekly is better. Update instructions, replace stale documents, and add what the last few weeks taught you. Treat it like an employee you talk to regularly.
3. Correct wrong memory plainly. "That is no longer true; please update it." Accurate memory helps you; stale memory misleads.
4. Use Handover blocks. When chats fill, carry the context forward with the Handover Skill rather than starting cold.
5. Close sessions deliberately. End substantial working sessions with: "What should we remember from this?" and save the answer into the project.

Repetition without capture is Level 2 work forever. Capture is what makes the partnership compound.

Habit Guide 4: Shape Your Stack

Keep your toolset current and fit for you.

Your stack is the set of AI tools you work with. The habit is not collecting tools; it is deliberately keeping a small stack sharp: one primary partner deepened daily, a few supporting tools that earn their keep, and a regular look at what is new.

1. Anchor on your primary partner. Depth with one beats dabbling with ten. The MASTER Rule from Building Your Tech Stack (Section 2) governs everything here.
2. Experiment on schedule, not on impulse. Set aside 30 minutes a week to test one new tool, model, or feature against your primary partner. If it clearly wins for a real task, note it; if not, move on.
3. Add only against a real need. Use the readiness checklist in Building Your Tech Stack before anything new enters the stack.
4. Prune. Anything you have not used in a month comes out. An unused subscription is not capability; it is clutter and cost.

Review your stack as part of the Weekly Review; that keeps this habit deliberate rather than compulsive.

Habit Guide 5: Engage Myriad Minds

Summon diverse perspectives; break out of echo chambers.

One mind, however good, has blind spots; and one AI reflecting one mind can amplify them. That is the Human Bias quadrant of the Bias Matrix. This habit is the standing defence: deliberately widening the circle of perspectives before you commit.

1. Install the Skill. The Engage Myriad Minds Skill convenes the council on command; use it for any real decision.
2. Use personas for pressure-testing. Reviewing a high-stakes proposal, have your AI assume a sceptical CFO, a visionary designer, and minds like Marcus Aurelius, and let them debate your plan.
3. Always ask for the opposing case. Before any big call: "Argue against this as strongly as you can." If the counter-argument is weak, proceed with confidence; if it is strong, you just saved yourself.
4. Cross-verify what matters. Run key facts and big conclusions through a second model or independent source; the standing remedy for AI Bias.
5. Include the humans. The habit is not only digital. Take what the council surfaces to your mentors, your team, and your community; the Four Critical Success Factors live here too.

The goal is never consensus; it is seeing the whole board before you move.

The Habit Builder

Design a habit properly, then build it with your AI partner until it sticks.

Most habits fail because people rely on motivation, start too big, and have no plan for the day they slip. The Habit Builder fixes the design, then does the thing no book can: it puts a partner alongside you who holds the habit, tracks it, troubleshoots when it slips, and keeps you going.

Get the Skill: download the pre-packaged habit-builder.skill file from the portal, or copy the raw instructions to paste into Claude's Custom Instructions or Project Knowledge. ▸ portal download

The Seven Steps of Habit Design

1. Name and Why. Be specific. Not 'exercise more,' but 'a ten-minute walk after lunch.' Vague habits fail; precise ones stick.
2. Shrink It. Make the first version almost embarrassingly small – e.g. Two push ups a day, one sentence. Pattern first, size later.
3. Anchor It. Attach it to an existing routine you already do. 'After I pour my morning coffee, I write three lines.'
4. Shape Environment. Make the good action easy and obvious; make competing distractions harder. Design beats discipline.
5. Make Progress Visible. Decide how you will see the chain building; a simple tick each day. What gets seen gets sustained.
6. Plan for the Miss. One miss is data, not failure; the rule is never miss twice. Agree with your partner what you do the day after.

Skill Code Preview

---
name: habit-builder
description: >
Guides the user through designing any habit properly and then building it with
their AI partner over time until it sticks. Use this skill whenever the user wants
to start, build, or fix a habit; signals include "help me build a habit", "I want
to start doing X", "I keep failing to stick to Y", "help me make X a habit", "I
can't get consistent with Z", or any request about forming, breaking, or sustaining
a behaviour. Grounded in habit-formation science (cue, routine, reward; small
actions; environment design; identity), expressed in original language; it does not
reproduce any proprietary habit framework. British English, no em-dashes, one
question at a time. Aligned to the Habits element of The Bolt Method v3.1.
---
# The Habit Builder
## Purpose
Help the user design a habit that will actually stick, then set up the ongoing
practice of building it with their AI partner. Most habits fail through poor design
and no follow-through, not weak character. This skill fixes the design and uses the
partnership to sustain it.
## When to use
Trigger when the user wants to start, build, sustain, or fix a habit, however phrased.
Do not trigger for one-off tasks; a habit is a repeated behaviour, not a single action.

The Overwhelm Buster

Break through overwhelm and return to presence, clarity, and focused action.

This protocol helps you break through overwhelm, confusion, and energy gridlock by returning to presence, clarity, and focused action. It can be used while walking, cleaning, pacing, driving, or moving around; anywhere you feel restless or overloaded but want to move forward. Part of the human development strand of The Bolt Method: the You component working alongside your AI partner.

When to use it. If you are feeling overwhelmed, scattered, or unsure where to begin, just say:

• "I'm feeling overwhelmed and don't know where to start."
• "Help me prioritise. I have too much going on."
• "I'm feeling wired but stuck. Can we clear some space?"

Your AI will walk you through a short four-step process. The goal is not to fix everything; it is to choose one area to unlock and take aligned action on, even while in motion.

The Four Steps

1. Focus: check in.

Slow things down. Before deciding what to do, name how you are actually feeling, in one or two words: wired, tired, pressured, frozen, hyper, frustrated. Naming the state is the first step out of it.

1. Map: three zones of focus.

Name just three key areas; not to-do items, but big themes taking up the most mental and emotional space right now. They might sound like "fundraising pressure", "family logistics", "client deadlines", or "financial stress". Quick labels, no overthinking.

1. Unlock: choose one.

If you could unlock just one of those areas, right here, right now, which would give you the most relief, clarity, or momentum for the rest of the day? Then, within that zone: one small step. Not the whole thing; just the next clean move, doable in the next 30 to 60 minutes.

1. Move: action and reset.

That is your zone, and that is your next move. You do not need to solve the whole thing; you just need to move the needle. Once it is done, come back and reflect, or simply close the loop and carry on. You have already reclaimed your power. Optionally, ask your AI to hold you to it and follow up later.

Installing It

The Overwhelm Buster runs from a short instruction script you give your AI.

The World-Class Business Consultant

Stop asking AI to summarise. Do this instead.

When you give your AI a document, a report, a transcript, or an article, the default request is "summarise this". A summary compresses information; it rarely creates value.

These eight prompts turn the same material into strategic insight, action, and leverage. Use them individually, or in sequence on the same document.

1. Extract Strategic Insights

AI becomes your McKinsey analyst.

1. Turn Information Into Action

AI becomes your project manager.

1. Surface Hidden Assumptions

AI becomes your devil's advocate.

1. Compare Opposing Views

AI becomes your debate moderator.

1. Distil for a Specific Role

AI becomes your specialist advisor.

1. Build a Reusable Model

AI becomes your systems architect.

1. Extract Contrarian Takeaways

AI becomes your thought provocateur.

1. Identify Leverage Points

AI becomes your force multiplier.

Most people use 5% of AI's capability. These prompts unlock the other 95%.

Ikigai Start With Why

Aligning your reason for being with the first principle of strategic action.

Ikigai is the Japanese art of finding your reason for being; the intersection of what you love, what you are good at, what the world needs, and what you can be paid for. In The Bolt Method, we align this deep self-discovery with the first principle of Thinking: Start With Why.

Get the Skill: download the pre-packaged ikigai-start-with-why.skill file from the portal, or copy the raw instructions into Claude's Custom Instructions or Project Knowledge. ▸ portal download

What Is Ikigai?

Ikigai (ee-key-guy) is a Japanese concept that combines iki, meaning 'alive' or 'life,' and gai, meaning 'benefit' or 'worth.' Together, they describe that which gives your life worth, meaning, or purpose.

The concept evolved from traditional Japanese medicine, which holds that physical wellbeing is deeply connected to one's mental and emotional health, and to one's sense of purpose. Neuroscientist Ken Mogi describes ikigai simply as "a reason to get up in the morning" or, more poetically, "waking up to joy."

In Okinawa, one of the world's Blue Zones where people live exceptionally long lives, a strong sense of ikigai is considered central to longevity and wellbeing.

The Four Elements

Your ikigai sits where all four overlap. Working with your AI partner, explore each element honestly, then look for the intersections; the result feeds directly into the Start with Why Skill as the deepest possible answer to "why does this matter to you?"

Skill Code Preview

The Weekly Review

Let your AI do the thinking; you make the calls.

Most weekly reviews fail because they make you do all the work. This one flips it. Your AI partner reviews the week first, makes the observations, and hands you clear choices to confirm or correct. You stay in charge of the judgement; it carries the load. Thirty minutes, mostly tapping options, ending with a short recommendations report you can act on.

Get the Skill: download the pre-packaged weekly-review.skill file from the portal, or copy the raw instructions into Claude's Custom Instructions or Project Knowledge. ▸ portal download

The Five Steps

1. AI Reviews the Week First. Drawing on your conversations, goals, and previous reviews, your partner tells you what it saw; what went well, what stalled, what it noticed. You are reacting to a first draft, not facing a blank page.
2. You Confirm or Adjust. After each observation it gives you multiple-choice options; agree, adjust, add, or 'that's not quite right.' You keep the judgement; it does the recall and the framing.
3. Set the Week Ahead. A simple, powerful frame; what to start doing, what to stop, what to keep going. Your partner proposes; you decide.
4. Choose One Habit Focus. Pick one of the five Habits of The Bolt Method to focus on for the week ahead, and commit to at least thirty minutes of daily AI practice.
5. Recommendations Report. A short, clear summary of the review and the week ahead, ready to act on and to revisit next time.

Start Your Weekly Review

Paste this into Claude to begin your AI-led review. Your partner will do the heavy lifting from the start.

The Five Habits to Choose From

Each week, pick one to go deeper on. Depth in one beats dabbling in all five.

• Walk and Talk. Working with AI by voice, often in motion.
• Make It Personal. Building the context and continuity that let your partner truly know you.
• Maintain Memory. Capturing and reusing what works, so the partnership compounds.
• Shape Your Stack. Keeping your AI toolset current and fit for you.
• Engage Myriad Minds. Summoning many perspectives to widen your thinking.

SECTION 4: CLAUDE USER GUIDES · THE PROGRESSION MAP

From beginner to advanced, mapped against the Five Levels of Collaboration.

Bolt Practice maintains a full library of Claude user guides in the portal, organised into three tiers: Beginner, Intermediate, and Advanced. Each tier maps broadly onto the Five Levels of Collaboration, though the mapping is a guide rather than a rule; features are flexible, and a skilled practitioner will use tools from every tier at every level. You do not need any of the Intermediate or Advanced material to complete this foundation programme. It is shown here so you can see the progression that is coming, and where each capability fits as your practice develops.

Beginner Guides (Tools level of collaboration)

The foundations. These are the features you will use every day, and the ones this foundation programme works with directly.

• Claude Settings. Your profile, personal preferences, and model choices; where you installed the Bolt Method preferences block.
• Claude Chats & Projects. Organising conversations and building Projects as context-rich homes for each area of your life and work.
• Claude Voice. Working by voice, both voice-to-text and voice-to-voice; the Walk and Talk habit in practice.
• Claude Search. Finding what you need, across your past conversations and live information.
• Claude Skills. Installing reusable processes, such as Start With Why, the Results Cycle, and Bolt to Result, so Claude runs them on demand.

Intermediate Guides (Agents level and higher)

Where Claude begins working for you, not just with you.

• Claude Connectors. Linking Claude to your email, calendar, documents, and business tools, so it works with your real data.
• Claude Plugins. Packaged bundles of skills and connections for specific roles and workflows.
• Claude Desktop App. Claude on your computer, with deeper access to your files and applications.
• Claude Co-Work. Agentic working sessions where Claude takes on multi-step tasks alongside you.
• Claude Scheduler. Recurring tasks that run on a schedule, without you prompting them.
• Claude Dispatch. Delegating work to background agents that report back when done.
• Claude Artefacts. Interactive, shareable outputs: documents, tools, and pages Claude builds for you.
• Claude Design. Visual and branded design work within Claude.

Advanced Guides (Partnership and Network levels)

The frontier: orchestrating agents and building systems.

• Claude Code. The most powerful way to work with Claude: agentic execution on your own computer, capable of multi-step projects, automation, and orchestrating teams of sub-agents.
• Other advanced platform use. Advanced configurations and multi-platform working patterns for Network-level practice.

Recommended Tech Stack Beyond Claude

Three tools we recommend alongside Claude, each covered by its own guide in the portal. As the Building Your Tech Stack guide sets out: add one at a time, and only when your Claude practice is established.

• NotebookLM. Research and document analysis across large collections of source material.
• Wispr Flow, or your device's built-in voice dictation. Fast, natural voice input for working with Claude at the speed of speech.
• Plaud, or another notetaker app or device. Capturing meetings and conversations so they can feed your AI partnership.

This is not a comprehensive list of everything covered in the portal; these are the main tools and features taught in the programme.

User Guide: Claude Search

Real-time web search integrated directly into Claude's reasoning.

Claude Search brings the live web into your Claude conversations. When you need current information, Claude can search the internet in real time and combine what it finds with its own reasoning and analysis capabilities. It is the fastest way to get up-to-date, intelligently analysed information without leaving your Claude workflow.

Getting Started

What is Claude Search?

Claude Search is Claude's built-in web search capability. When enabled, Claude can search the internet in real time to find current information, verify facts, and supplement its responses with up-to-date data from the web.

Unlike Perplexity (which is a dedicated search-first tool), Claude Search is integrated directly into Claude's conversational interface. This means you get Claude's superior reasoning and analysis capabilities combined with access to current web information, all in one place.

Web search is available across Claude plans, including the free tier. This was accurate as of 15 July 2026; check claude.ai for the current position.

How to Enable and Use Claude Search

1. Go to claude.ai and sign in.
2. Start a new conversation. Look for the web search option in the chat input toolbar.
3. Enable web search for your query.
4. Type your question and send it. Claude will search the web before responding.
5. Claude will indicate when it has used web search and may include citations to the sources it found.

Alternatively, Claude will automatically use search when it determines that current information is needed, even without you explicitly enabling it.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

Example Claude Search Queries

Key Use Cases

• Real-Time Fact Checking. Ask Claude to verify facts, statistics, and claims against current web sources mid-conversation, without switching to another tool.
• Current Market Data. Get up-to-date pricing, market statistics, and industry data combined with Claude's analysis and recommendations.
• News and Events Research. Research recent news, events, and developments and have Claude immediately apply its reasoning to the implications for your business.
• Competitor Monitoring. Ask Claude to search for recent news or updates about specific competitors and analyse what it finds.
• Product Research. Research current product specifications, pricing, and reviews for purchasing decisions, with Claude helping you compare and decide.
• Regulatory Updates. Stay current on regulatory changes in your industry by asking Claude to search for recent developments and explain their implications.

Bolt Method Alignment

Claude Search supports the Test for Truth principle from The Bolt Method. It allows you to verify Claude's outputs against current, live web sources without leaving the conversation, ensuring that the information underpinning your decisions is accurate and up to date.

User Guide: Claude Skills

Install reusable AI behaviours into your Claude Projects.

Skills are the building blocks of a high-performance AI partnership. By installing Skills into Claude, you give it a persistent, structured way of working that activates automatically when you need it, without having to re-explain your process every time. For SMEs, Skills are the fastest way to systematise your best practices and embed them permanently into your AI partner.

Getting Started

What are Claude Skills?

Skills are pre-built, reusable instruction sets that you install into Claude to give it a specific, persistent capability. A Skill tells Claude how to behave in a particular context, what process to follow, and how to format its outputs.

Skills are stored as .skill files and can be shared across teams and projects. The Bolt Method portal includes several Skills you can download and install directly.

The key difference between a Skill and a regular prompt is persistence: once installed, a Skill is always available, in every conversation and project. You never have to re-explain the process. Claude just knows.

Installing a Skill into Claude

1. Download the .skill file from this portal. It is a small ZIP package containing the skill's instructions.
2. In Claude, go to Settings, then Customize, then Skills. Make sure code execution and file creation is enabled in your settings; Skills depend on it.
3. Click the "+" button, then Create skill, and upload the .skill file.
4. The Skill appears in your Skills list; check its toggle is on. It is now available across your conversations, and custom skills stay private to your account.
5. Test it by triggering the Skill with its defined phrase (e.g. "Run the Results Cycle on this brief.").

You can install as many Skills as you like; Claude activates the right one from the context of your request. For best results, give each Skill a distinct trigger phrase so there is never any doubt which one you mean. Skills are available on Free, Pro and Max plans, and on Team and Enterprise plans where an owner has enabled them; accurate as of July 2026.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

Triggering a Skill

Each Skill has defined trigger phrases. For example:

• Run the Results Cycle on this project brief.
• Run Engage Myriad Minds on this decision.
• Run my Weekly Review.
• Run the Sales Call Debrief on this transcript.

You can also trigger a Skill implicitly by describing the situation, and Claude will recognise when to apply it based on the context you provide.

Building Your Own Skills

You can create your own custom Skills for any repetitive process in your business. A Skill is simply a well-structured set of instructions packaged in a small file Claude can install.

To build a Skill:

1. Identify a process you repeat regularly (e.g. writing proposals, running debriefs, creating reports).
2. Write out the steps Claude should follow, the questions it should ask, and the format of the output.
3. Test the instructions in a Claude conversation and refine until the output is consistently excellent.
4. Ask Claude to package the instructions as a skill file (a folder containing a SKILL.md instruction file, zipped), then install it via Settings, Customize, Skills.

Claude can help you write your own Skills. Describe a process you do regularly and ask Claude to turn it into a structured, reusable Skill file.

Bolt Method Alignment

Skills are a direct implementation of the Maintain Memory habit from The Bolt Method. They allow you to capture and reuse what works, so the partnership compounds over time rather than starting from scratch in every conversation. Every Skill you build is an asset that makes your AI partnership more valuable with each use.

12 Ways to Use Skills in an SME Business

Each of these use cases represents a Skill you can build once and use indefinitely. The more consistently your team uses them, the more valuable they become, creating a compounding advantage over businesses that start from scratch every time.

1. Sales Call Debrief. After every sales call, trigger a Skill that extracts: the prospect's pain points, objections raised, commitments made, and recommended next steps. Every rep follows the same structured debrief automatically.
2. New Client Onboarding. Install an onboarding Skill that walks Claude through a structured welcome sequence: gathering client context, setting expectations, generating a personalised onboarding plan, and creating the first project brief.
3. Weekly Management Report. A Skill that compiles a consistent weekly management report from your inputs: key wins, challenges, metrics, decisions made, and priorities for the coming week. Same format every time, ready in minutes.
4. Proposal Writing. A proposal Skill that takes a client brief and generates a structured, on-brand proposal covering: executive summary, scope of work, methodology, timeline, investment, and next steps.
5. Competitor Analysis. A research Skill that, when given a competitor name, systematically analyses: their positioning, pricing, product strengths and weaknesses, recent news, and strategic implications for your business.
6. Customer Complaint Resolution. A Skill that guides Claude through a structured complaint resolution process: acknowledging the issue, identifying the root cause, proposing a resolution, and drafting a professional response email.
7. Job Interview Preparation. A Skill that prepares candidates for interviews by generating role-specific questions, coaching on STAR-format answers, identifying likely competency areas, and producing a pre-interview briefing document.
8. Monthly Financial Review. A Skill that structures a monthly financial review: revenue vs target, cost analysis, cash flow commentary, key variances, and a three-point action plan for the following month.
9. Content Calendar Planning. A Skill that takes your key messages and audience and generates a structured monthly content calendar with topic ideas, formats, platforms, and calls to action for each piece.
10. Risk Assessment. A Skill that applies a consistent risk assessment framework to any project, decision, or initiative: identifying risks, rating likelihood and impact, suggesting mitigations, and producing a structured risk register.
11. Team Meeting Facilitation. A Skill that structures team meetings: generating a focused agenda from your inputs, capturing decisions and actions during the meeting, and producing a formatted minutes document with owners and deadlines.
12. Strategic Planning Session. A Skill that facilitates a structured strategy session: clarifying the vision, analysing the current position, identifying strategic options, stress-testing assumptions, and producing a one-page strategic summary.

User Guide: Claude Plugins

Ready-made bundles of Skills and Connectors for whole roles and workflows.

Plugins are packaged bundles of Skills, Connectors, and specialist sub-agents that set Claude up for a complete role or workflow in one installation. Where a Skill teaches Claude one process and a Connector links it to one service, a plugin delivers a ready-to-go setup: install a sales plugin, for example, and Claude arrives with call preparation, pipeline review, and outreach skills, plus connections to the tools that work needs.

Getting Started

What are Claude Plugins? Claude offers a growing library of plugins for common knowledge work: sales, marketing, finance, legal, HR, operations, design, data analysis, and more. Plugins are available on paid Claude plans, and the core library is included at no extra cost. Accurate as of July 2026; check claude.com/plugins for the current library.

Installing a Plugin

1. Browse the plugin library at claude.com/plugins, or in Claude's Customize settings.
2. Click Install on the plugin you want.
3. Connect any services the plugin's connectors need; you will be asked to authenticate with each one.
4. The plugin's skills are then available in chat on the web, in Claude Desktop, and in Co-Work. Some advanced components (sub-agents and hooks) run only in Co-Work.

You can also upload a custom plugin file if you have built one, or received one from a colleague. Claude asks for your confirmation before a plugin takes consequential actions, such as sending, publishing, or changing data in a connected service; review what it intends before approving.

Your own plugin. As your practice matures, a team can package its own working methods as a custom plugin: your Skills, your Connectors, one install for every new joiner. This is where systematising your best practices pays its dividend.

Bolt Method Alignment

Plugins are the Shape Your Stack habit applied at the level of a whole role, and Maintain Memory at team scale: the working methods of your best people, captured once and installed everywhere.

User Guide: Claude Connectors

Connect Claude directly to your tools and data sources.

Claude Connectors allow Claude to read from and write to external services in real time, eliminating the need to copy and paste data. With Connectors, Claude becomes a genuine operating partner across your entire digital workspace.

Getting Started

What are Claude Connectors?

Connectors are integrations that allow Claude to read from and write to external services and data sources in real time. Rather than copying and pasting data into Claude, Connectors let Claude access your tools directly during a conversation.

Examples include: Google Drive, Notion, GitHub, Jira, Confluence, Salesforce, and custom API endpoints.

Connectors are available across Claude plans, including a limited allowance on the free tier; exactly which connectors are included varies by plan and changes as the platform evolves. Accurate as of July 2026; check claude.ai for the current position.

Setting Up a Connector

1. In Claude, go to Settings, then Customize, then Connectors.
2. Browse the connector directory and click Connect on the service you want (Google Drive, Gmail, Notion, and many more). You will be asked to sign in to that service and grant access.
3. On Team and Enterprise plans, an organisation owner may need to enable the connector first in Organization settings.
4. Advanced: you can also add a custom connector by URL, linking Claude to specialist or in-house tools.

Using a Connector in a Conversation

1. Click the "+" button at the lower left of the chat input and choose Connectors.
2. Toggle on the connectors you want available for this conversation.
3. Ask naturally: "Find the latest version of the Henderson proposal in my Drive and summarise what changed." Claude asks for your confirmation before writing to or changing anything in a connected service.

Key Use Cases

• Document Search. Ask Claude to find and summarise documents across Google Drive, Notion, or Confluence without leaving the chat.
• Project Management. Pull open tasks from Jira or Asana and ask Claude to prioritise, summarise, or draft updates.
• Code Review. Connect GitHub and ask Claude to review pull requests, explain diffs, or suggest improvements.
• CRM Intelligence. Connect Salesforce and ask Claude to summarise a customer's history before a call.

Bolt Method Alignment

Connectors operationalise the Maintain Memory and Shape Your Stack habits from The Bolt Method. They extend Claude's context beyond the chat window, allowing it to act as a genuine partner across your entire digital workspace rather than an isolated tool.

SECTION 5: RECOMMENDED TECH STACK

The tools we recommend alongside Claude. Add one at a time, as the Building Your Tech Stack guide sets out.

Tool Guide: NotebookLM

Your private AI research assistant, grounded entirely in your own sources.

NotebookLM is one of the most powerful and underused AI tools available. It creates a private AI research assistant from your own documents, videos, audio files, and websites, giving you cited, accurate answers based exclusively on your material. Grounded and cited, so every claim can be checked against your sources; no mixing with general internet knowledge. Just deep, reliable intelligence from the sources you trust.

Getting Started

What is NotebookLM and Why is it Different?

NotebookLM is Google's AI-powered research and knowledge management platform. The fundamental difference between NotebookLM and general AI tools like Claude or ChatGPT is this: NotebookLM only uses the sources you provide. It never draws on its general training data to answer your questions.

This makes it exceptionally reliable for research, document analysis, and knowledge management. Every answer it gives is grounded in your specific material, with citations linking directly to the exact passage in your source. Every claim is grounded and cited, so you can check it against your sources, and your content is never mixed with general internet knowledge.

Think of it as creating a private, intelligent research assistant that knows only what you have taught it, and can prove every claim it makes.

NotebookLM is free to use with a Google account. NotebookLM Plus (paid) offers higher usage limits, more sources per notebook, and additional Studio features. Plan details accurate as of 15 July 2026; check notebooklm.google.com for the current position.

Creating Your First Notebook

1. Open your browser and go to notebooklm.google.com.
2. Sign in with your Google account. If you do not have one, create a free account at google.com.
3. Click the + New Notebook button in the top-left corner.
4. Give your notebook a clear, descriptive name (e.g. "Q3 Competitor Research" or "Sales Training Materials").
5. You will land on the main notebook interface. This is where you will add your sources.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

The Three Panels: Understanding the Interface

The NotebookLM interface is divided into three distinct panels, each with a specific purpose.

Panel 1: Sources (Left Panel)

The Sources panel is where you manage all the documents, files, and links that form your notebook's knowledge base. Everything NotebookLM knows comes from this panel.

• Add sources by clicking the + Add Source button.
• Select which sources to use by ticking or unticking the checkbox next to each source. This is powerful: you can have 20 sources in a notebook but only query 3 of them at a time.
• View source details by clicking on any source to see a summary and key topics.
• Delete sources by hovering over a source and clicking the three-dot menu.

Use source selection strategically. If you are researching a specific topic, deselect unrelated sources to get more focused, accurate answers.

Panel 2: Chat (Centre Panel)

The Chat panel is your primary interface for asking questions and getting answers from your sources. It works like a conversation, but every response is grounded in your selected sources.

• Ask questions in plain English about your sources.
• View citations by clicking the numbered references in each response, which jump directly to the relevant passage in your source.
• Save responses as notes by clicking the pin icon, which saves them to your Notes panel.
• Generate structured outputs using the quick-action buttons above the chat input: Study Guide, Briefing Document, FAQ, Timeline, and Table of Contents.

Panel 3: Studio (Right Panel)

The Studio panel is where NotebookLM generates polished, shareable outputs from your sources. It is the production side of the tool.

• Audio Overview: generate a podcast-style conversation between two AI hosts discussing your sources. Excellent for absorbing dense material while commuting or walking.
• Customise the Audio Overview: before generating, you can give instructions to the AI hosts: "Focus on the practical implications", "Keep it under 10 minutes", or "Explain it as if to a non-expert".
• Custom branding: on NotebookLM Plus, you can customise the Audio Overview with your own intro and outro, making it suitable for sharing with clients or participants under your brand.
• Notes: all saved chat responses and manually written notes appear here, forming a structured knowledge base from your research.

All Source Types: What You Can Upload

NotebookLM accepts a wide range of source types.

• PDF Files. Upload research papers, reports, contracts, manuals, and any PDF document. NotebookLM reads the full text and can answer questions about specific sections.
• Google Docs. Connect directly to Google Docs in your Drive. NotebookLM syncs with the live document, so updates to the Doc are reflected in your notebook.
• Google Slides. Upload presentations from Google Slides. NotebookLM reads the slide content and speaker notes.
• Websites & Web Pages. Paste any public URL and NotebookLM will read and index the page content. Useful for competitor websites, news articles, or documentation pages.
• YouTube Videos. Paste a YouTube URL and NotebookLM will read the video's transcript (where available). Ask questions about what was said in the video.
• Audio Files. Upload MP3, WAV, or other audio files. NotebookLM transcribes the audio and makes it searchable and queryable. Ideal for recorded meetings, interviews, or podcasts.
• Text Files (.txt). Upload plain text files, including transcripts, notes, or any text-based content.
• Copied Text. Paste text directly into NotebookLM as a source. Useful for short documents, email threads, or content you have copied from other tools.

You can add up to 50 sources per notebook on the free plan, with a higher limit on Plus. Mix and match source types freely; NotebookLM handles them all equally.

Selecting Sources for Focused Queries

One of NotebookLM's most powerful and underused features is selective source querying. Rather than always querying all your sources, you can choose exactly which ones the AI uses to answer a specific question.

How to select sources:

1. In the Sources panel on the left, you will see a checkbox next to each source.
2. Tick only the sources you want to query for your current question.
3. The chat will now only draw answers from your selected sources.
4. Deselect all and reselect different sources to shift your query focus.

Why this matters: if you have a notebook with 20 competitor research documents, you can select just three competitors to compare them directly, without the other 17 sources diluting the response.

Tip: create a single large notebook for a broad topic (e.g. "AI Market Research") and use source selection to dynamically focus your queries rather than creating multiple smaller notebooks.

Studio: Audio Overviews, Custom Branding, and Outputs

The Studio panel is where NotebookLM transforms your research into polished, shareable content.

Audio Overview (Podcast Generation)

The Audio Overview generates a natural, engaging podcast-style conversation between two AI hosts who discuss and debate the content of your sources. It is one of the most impressive AI features available anywhere.

• Generate: click Generate in the Studio panel. The process takes 1-3 minutes depending on source volume.
• Customise before generating: click Customise and give instructions: "Focus on the business implications", "Keep it under 8 minutes", "Explain it for a non-technical audience", or "Emphasise the risks and challenges".
• Listen and share: play the audio directly in NotebookLM, download the MP3, or share a link.
• Interactive mode: on some versions, you can interrupt the podcast and ask the hosts questions in real time.

Custom Branding (NotebookLM Plus)

NotebookLM Plus allows you to brand your Audio Overviews with custom intros and outros, making them suitable for sharing with clients, participants, and audiences under your own brand.

• Custom intro: record or upload a branded intro (e.g. "Welcome to the Bolt Practice AI Briefing...") that plays before the AI hosts begin.
• Custom outro: add a branded sign-off or call to action at the end.
• Use case: create branded weekly AI briefings, client research summaries, or training audio that sounds professionally produced.

Structured Document Outputs

Using the quick-action buttons in the Chat panel, NotebookLM can generate these structured outputs from your sources:

• Study Guide: key concepts, definitions, and review questions.
• Briefing Document: a concise executive summary of all sources.
• FAQ: the most likely questions and answers based on your sources.
• Timeline: chronological events extracted from your sources.
• Table of Contents: a structured overview of the topics covered.

Advanced Chat Techniques

Go beyond basic questions with these advanced chat approaches:

Key Use Cases

• Competitor Research. Upload competitor websites, annual reports, and press releases. Query them selectively to compare positioning, pricing, and product strategy across multiple competitors simultaneously.
• Contract and Legal Review. Upload contracts, terms, and legal documents. Ask specific questions to find clauses, obligations, and risks without reading every page manually.
• Meeting Preparation. Upload briefing documents, background research, and previous meeting notes. Ask NotebookLM to prepare you with key points, likely questions, and talking points.
• Training and Onboarding. Upload training materials, process documents, and company policies. New team members can ask questions and get cited answers from the official documentation.
• Research Synthesis. Upload multiple research papers or industry reports and ask NotebookLM to synthesise the key findings, identify consensus, and surface contradictions across all sources.
• Branded Audio Briefings. Create weekly AI-generated audio briefings from curated sources, branded with your intro and outro, and share them with clients or participants as a premium value-add.
• Due Diligence. Upload investment documents, pitch decks, and financial reports for a company. Ask structured due diligence questions and get cited answers from the actual documents.
• Content Research. Upload source material for an article, report, or presentation. Ask NotebookLM to surface the most relevant quotes, statistics, and arguments from your research.

Bolt Method Alignment

NotebookLM is the ideal tool for the Gather stage of the Results Cycle from The Bolt Method. It ensures you build your understanding on verified, cited sources before reasoning and deciding, directly operationalising the Test for Truth principle by grounding every answer in evidence you can verify.

Tool Guide: Built-in Voice Dictation

Free voice dictation built into every Mac and Windows computer.

Every Mac and Windows computer comes with built-in voice dictation that is free, requires no additional software, and works in any text field. It is the fastest way to get started with voice input, and for many people it is all they need. This guide covers how to activate it, set it up, and troubleshoot it on both platforms.

Mac: How to Enable and Use Voice Dictation

The keyboard shortcut: press the Fn key twice (or Fn + F5 on some keyboards) to activate dictation. A microphone icon will appear. Speak, then press Fn again or press Return to stop.

If it is not working, set it up:

1. Click the Apple menu in the top-left corner of your screen.
2. Select System Settings (or System Preferences on older Macs).
3. Click Keyboard.
4. Scroll down to find Dictation and toggle it On.
5. Choose your language from the dropdown menu.
6. Optionally, change the shortcut key if you prefer a different activation method.
7. Close Settings. Test by clicking into a text field and pressing Fn twice.

Offline dictation: on older Macs, look for Enhanced Dictation in the Dictation settings to download a local language model, allowing dictation to work without an internet connection and with lower latency. On newer Macs, on-device processing happens automatically.

On newer Macs with Apple Silicon (M1, M2, M3, M4), dictation is exceptionally fast and accurate. If you have an older Intel Mac, consider upgrading to Wispr Flow for better performance.

Windows: How to Enable and Use Voice Dictation

The keyboard shortcut: press Windows key + H to open the dictation toolbar. Click the microphone button and speak. Click it again or say "Stop dictation" to stop.

If it is not working, set it up:

1. Press Windows key + I to open Settings.
2. Click Accessibility in the left sidebar.
3. Scroll down and click Speech.
4. Toggle Windows Speech Recognition to On.
5. Follow the setup wizard to configure your microphone and train the speech model to your voice.
6. Close Settings. Test by pressing Windows + H in any text field.

Voice Access (Windows 11): Windows 11 includes a more advanced feature called Voice Access, which allows you to control your entire computer by voice, not just dictate text. Enable it in Settings, under Accessibility, then Speech, then Voice Access.

Windows dictation works best with a good quality microphone. If you are using a laptop's built-in microphone, consider a USB or Bluetooth headset for significantly better accuracy.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

Voice Commands for Dictation

Both Mac and Windows dictation support voice commands for formatting and editing:

Troubleshooting: When Dictation is Not Working

• Microphone not detected. Check that your microphone is plugged in and selected as the input device. On Mac: System Settings, then Sound, then Input. On Windows: Settings, then System, then Sound, then Input.
• Dictation activates but produces no text. Check your microphone volume. On Mac, look at the input level bar in Sound settings; it should move when you speak. On Windows, check the microphone level in Sound settings.
• Poor accuracy or wrong words. Speak more clearly and at a moderate pace. Reduce background noise. On Windows, run the speech training wizard again.
• Shortcut key not working. Another application may be using the same shortcut. On Mac, check System Settings, then Keyboard, then Keyboard Shortcuts for conflicts. On Windows, check if any other software intercepts Windows + H.
• Dictation stops after a few seconds. This is a timeout feature. Speak continuously without long pauses, or disable the timeout in dictation settings.

Key Use Cases

• Quick Claude Prompts. Activate dictation in the Claude chat field and speak your prompt instead of typing it, saving time on shorter queries.
• Email Drafting. Dictate email replies quickly without typing, particularly useful for mobile-style responses on a desktop.
• Note Taking. Dictate notes, observations, and ideas directly into Notion, Word, or any note-taking app.
• Form Filling. Dictate text into web forms, search fields, and other input areas across any application.
• Accessibility. For users with typing difficulties, built-in dictation provides a free, always-available alternative to keyboard input.
• First Draft Writing. Speak a first draft of a document or email, then refine it with Claude or manually.

Bolt Method Alignment

Built-in voice dictation is the entry point to the Walk and Talk habit from The Bolt Method. Start here to experience voice-first working before investing in more advanced tools like Wispr Flow. The habit of speaking your thoughts rather than typing them is transformative for AI partnership.

Tool Guide: Wispr Flow

AI-powered voice dictation across your entire computer.

Wispr Flow transforms voice into clean, professional text across every application on your computer. It removes filler words, corrects grammar, adapts to context, and works everywhere from Claude to Gmail to Notion. For anyone who thinks faster than they type, or who wants to work more naturally and quickly, Wispr Flow is transformative.

Getting Started

What is Wispr Flow?

Wispr Flow is an AI-powered voice dictation tool that works across your entire computer. It transcribes your speech into text in any application, from email to documents to chat interfaces, with exceptional accuracy and speed.

Unlike basic dictation (like Apple's built-in dictation or Windows Speech Recognition), Wispr Flow uses AI to actively clean up your speech: removing filler words, correcting grammar, reformatting sentences, and adapting the tone and style to match the application you are using.

The result is that you can speak naturally, imperfectly, and quickly, and Wispr Flow produces clean, professional text as if you had typed it carefully.

Installing and Setting Up Wispr Flow

1. Visit the Wispr Flow website and click Download. Wispr Flow is available for Mac and Windows.
2. Run the installer and follow the on-screen instructions. On Mac, you may need to grant accessibility permissions in System Settings, under Privacy & Security, then Accessibility.
3. Create a Wispr Flow account and sign in.
4. Set your activation shortcut. The default is to hold the Option key (Mac) or hold the Alt key (Windows). You can customise this in Settings.
5. Test it: click into any text field (e.g. a new email or a Claude chat), hold your shortcut key, speak, and release. Your words will appear as clean text.

On Mac, Wispr Flow requires Accessibility permissions to work across all applications. If it is not working, go to System Settings, then Privacy & Security, then Accessibility, and ensure Wispr Flow is enabled.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

How to Use Wispr Flow

Using Wispr Flow is simple once it is set up:

1. Click into any text field in any application (Claude, Gmail, Notion, Slack, Word, etc.).
2. Hold your activation key (default: Option on Mac, Alt on Windows).
3. Speak naturally. You do not need to speak slowly or carefully; Wispr Flow handles natural, conversational speech.
4. Release the key when you have finished speaking.
5. Wispr Flow will process your speech and insert the cleaned, formatted text into the field.

For long dictations (more than a paragraph), hold the key and speak continuously. Wispr Flow processes the full speech when you release the key, so longer dictations are fine.

AI Cleanup and Intelligent Formatting

Wispr Flow does far more than basic transcription. Its AI actively improves your speech in real time:

• Filler Word Removal. Automatically removes 'um', 'uh', 'like', 'you know', and other filler words from your speech, producing clean, professional text.
• Grammar Correction. Corrects grammatical errors, tense inconsistencies, and sentence structure issues as it transcribes.
• Context-Aware Formatting. Adapts the tone and formatting to match the application. In Slack, it writes casually. In email, it writes formally. In Claude, it writes as a clear, structured prompt.
• Punctuation and Capitalisation. Adds appropriate punctuation, capitalises proper nouns, and formats lists and paragraphs correctly.
• Custom Vocabulary. Add technical terms, brand names, and jargon in Settings, under Custom Vocabulary, so Wispr Flow always transcribes them correctly.
• Multi-Language Support. Supports over 100 languages with high accuracy. Switch languages in Settings or speak in a different language and Wispr Flow will detect it automatically.

Using Wispr Flow with Claude

Wispr Flow and Claude are an exceptionally powerful combination. Instead of typing long, detailed prompts, you can speak them naturally and Wispr Flow will produce clean, structured text that Claude can work with immediately.

Example workflow:

1. Open Claude.ai and click into the chat input field.
2. Hold your Wispr Flow activation key and speak your prompt naturally.
3. Release the key. Wispr Flow will insert a clean, formatted prompt into Claude.
4. Press Enter to send.

This is particularly powerful for:

• Long, detailed prompts that would take minutes to type.
• Capturing thoughts quickly before they are lost.
• Working while moving (standing at your desk, pacing, etc.).
• Dictating meeting notes, ideas, or observations in real time.

Tip: combine Wispr Flow with Plaud for a complete voice-first workflow: Plaud captures your meetings, Wispr Flow lets you dictate your analysis and follow-up prompts into Claude, and Claude processes everything into structured outputs.

Key Use Cases

• Long Claude Prompts. Dictate detailed, nuanced prompts into Claude in seconds rather than typing them out, dramatically speeding up your AI workflow.
• Email Drafting. Dictate emails naturally and Wispr Flow produces clean, professional text ready to send, adapting the tone to match your email client.
• Meeting Notes. Dictate observations, decisions, and action items in real time during or immediately after meetings, without breaking your flow.
• Document Drafting. Dictate first drafts of documents, reports, and proposals at speaking speed, then refine with Claude.
• Slack and Teams Messages. Dictate messages in chat tools with Wispr Flow automatically adjusting the tone to be appropriately casual and conversational.
• Idea Capture. Capture ideas, insights, and observations instantly by speaking them, ensuring nothing is lost between thinking and recording.
• Multilingual Communication. Dictate in your native language and have Wispr Flow transcribe accurately, even for technical or domain-specific vocabulary.
• Accessibility. For anyone with repetitive strain injury, dyslexia, or other conditions that make typing difficult, Wispr Flow provides a natural, effective alternative.

Bolt Method Alignment

Wispr Flow directly enables the Walk and Talk habit from The Bolt Method. It removes the friction of typing, allowing you to think out loud and have Claude respond in a natural, conversational flow. The keyboard is no longer a bottleneck between your thinking and your AI partnership.

Tool Guide: Plaud

AI-powered voice recording, transcription, and meeting intelligence.

Plaud is the physical extension of your AI partnership. The Plaud Note device and app capture every conversation, meeting, and phone call with high-quality audio, then automatically transcribe, summarise, and analyse them using AI. Custom templates mean every sales call, coaching session, or team meeting produces a consistent, structured output automatically, with no manual note-taking required.

Getting Started

What is Plaud and What Does it Come With?

Plaud is an AI-powered voice recording ecosystem. It consists of two products that work together:

Plaud Note (physical device). A slim, credit-card-sized hardware device that attaches magnetically to the back of your iPhone or Android phone. It has a dedicated microphone optimised for capturing clear audio in meetings, interviews, and conversations. You press the physical button to start and stop recording. The device charges via USB-C and has a battery life of approximately 30 hours of recording.

Plaud App (iPhone & Android). The companion app that syncs with your Plaud Note device and also works as a standalone recorder on your phone. The app handles all the AI processing: transcription, summarisation, action item extraction, infographic generation, and Q&A. You can also record directly from the app without the physical device.

You do not need the physical Plaud Note device to use the app. The Plaud app works as a standalone voice recorder and AI processor on your phone, though the physical device provides superior audio quality and battery life for longer recordings.

Setting Up Plaud: Step by Step

Step 1: Get the App

1. Search for Plaud Note in the App Store (iPhone) or Google Play Store (Android).
2. Download and install the app. It is free to download; some AI features require a subscription.
3. Open the app and create a Plaud account using your email address.

Step 2: Connect the Physical Device (if you have one)

1. Attach the Plaud Note device to the back of your phone using the magnetic mount.
2. Open the Plaud app and tap Connect Device. The app will detect the Plaud Note automatically via Bluetooth.
3. Follow the on-screen pairing instructions. The device will vibrate once to confirm it is connected.

Step 3: Make a Test Recording

1. Press the button on the Plaud Note (or tap the record button in the app) to start recording.
2. Speak for 30 seconds, then press the button again to stop.
3. The recording will appear in the app and begin transcribing automatically.
4. Once transcription is complete, tap the recording to see the full transcript and AI summary.

Stuck at any point? Simply describe the problem to Claude and ask for help. For example: "I am trying to [step] but I am seeing [error/issue]. What should I do?" Claude will walk you through it step by step. If you are on a website, you can also use the Claude browser extension to let Claude see the page and help you fix or navigate it directly.

Recording Phone Calls

One of Plaud's most valuable features is the ability to record phone calls directly.

On iPhone:

1. Before your call, attach the Plaud Note device to the back of your phone.
2. Open the Plaud app and start a recording session.
3. Make or receive your phone call as normal. The Plaud Note's microphone will capture both sides of the conversation through the phone's speaker.
4. For best results, use speakerphone during the call so both voices are captured clearly.
5. Stop the recording when the call ends. The app will transcribe and process the recording automatically.

On Android:

Some Android devices support direct call recording through the Plaud app. Check the app settings under Call Recording to see if your device is supported. If not, use the speakerphone method described above.

Recording In-Person Meetings with Plaud Note

Plaud Note is designed for in-person meetings. Here is how to get the best results:

1. Place your phone (with Plaud Note attached) on the table, face up, in the centre of the group.
2. Press the button on the Plaud Note to start recording before the meeting begins.
3. The device will capture all voices in the room. For meetings with more than 6 people, consider placing the phone closer to the main speaker.
4. Press the button to stop recording when the meeting ends.
5. Open the Plaud app. The recording will sync and begin transcription automatically.

For video calls (Zoom, Teams, Google Meet), use the Plaud app on your phone to record the audio from your computer's speakers while the call is in progress, or use the screen recording feature on your computer and then upload the audio file to Plaud.

Custom Templates: Automate Your Recurring Recordings

Custom Templates are one of Plaud's most powerful features. They allow you to define exactly how the AI should process a specific type of recording, so you get a consistent, structured output every time without having to re-prompt.

How to create a Custom Template:

1. Open the Plaud app and go to Templates, then Create Template.
2. Give the template a name (e.g. "Sales Call", "Coaching Session", "Team Meeting").
3. Define the output structure: choose which sections you want in the summary (e.g. Key Points, Action Items, Decisions Made, Follow-up Required).
4. Add custom instructions: "Always identify the client's main objection", "Extract all commitments made by both parties", or "Summarise the coaching insights and the agreed actions".
5. Save the template. Next time you record this type of meeting, select the template before processing and Plaud will apply it automatically.

Example template structures:

AI Features: Transcription, Summaries, Infographics, and Q&A

Once a recording is processed, Plaud's AI gives you multiple ways to work with the content:

• Full Transcript. A complete, timestamped transcript of the recording with speaker identification (where audio quality allows). You can search the transcript, copy sections, and export it as a text file.
• AI Summary. A structured summary of the recording based on your template or the default format. Includes key points, action items, and decisions. Exportable as a formatted document.
• Infographic Generation. Plaud can generate a visual infographic summarising the key points of your recording. This is particularly useful for sharing meeting outcomes with people who were not present, or for creating visual summaries for presentations and reports.
• Q&A: Ask Questions About the Recording. After processing, you can ask the AI questions about the specific recording. For example: "What did the client say about their budget?", "What was the main objection raised?", or "What did we agree to do by Friday?" The AI answers from the recording transcript only.

Integrating Plaud with Claude

Plaud and Claude work exceptionally well together. Here is the workflow:

1. After a meeting, open the Plaud app and tap Export Transcript.
2. Copy the transcript text or export it as a .txt file.
3. Open Claude and paste the transcript with a specific prompt.

Tip: save your best Plaud-to-Claude prompts in a Claude Project as a Skill or in Project Knowledge, so you can trigger them instantly after every recording without rewriting the prompt each time.

Key Use Cases

• Sales Call Intelligence. Record every sales call and use a custom template to automatically extract pain points, objections, commitments, and next steps. Build a searchable library of every sales conversation.
• Coaching Session Records. Capture coaching sessions with a dedicated template that extracts client goals, breakthroughs, homework, and progress. Share formatted session notes with clients automatically.
• Team Meeting Minutes. Replace manual minute-taking with automatic transcription and structured summaries. Every team meeting produces a formatted record with action items, owners, and deadlines.
• Phone Call Recording. Record important phone calls and have them transcribed and summarised automatically. Never miss a commitment or detail from a verbal conversation again.
• Interview Transcription. Record candidate interviews, client interviews, or research interviews and get full transcripts with AI summaries, saving hours of manual transcription work.
• Infographic Reports. Generate visual infographic summaries of meetings and share them with stakeholders who were not present, making complex discussions easy to absorb at a glance.
• Due Diligence Calls. Record investor calls, reference checks, and due diligence conversations. Use Q&A to ask specific questions about what was said in any recording.
• Training and Onboarding. Record training sessions and onboarding conversations. Build a library of searchable, queryable recordings that new team members can access and ask questions about.

Bolt Method Alignment

Plaud operationalises the Walk and Talk and Maintain Memory habits from The Bolt Method. It captures the intellectual capital from every conversation, ensuring nothing is lost and everything can be processed, stored, and acted upon with Claude's help. The combination of Plaud for capture and Claude for analysis is one of the most powerful productivity workflows available.

SECTION 6: OPTIONAL TECH STACK

Specialist tools for specific needs. Explore these only when your core stack is established and you have a genuine use case (see Step 5 of Building Your Tech Stack).

The portal contains a full guide for each of these tools; this document lists them so you can see what is available as your practice develops.

• Perplexity. AI-powered search engine for fast, cited answers from the live web.
• ChatGPT Image. Image generation for illustrations, mockups, and visual content.
• Grok Imagine. xAI's image and video generation tool for creative content.
• HiggsField. AI video generation for cinematic and motion content.
• Lovable. Build working web applications from plain-English descriptions.
• Meet Alfred. LinkedIn and multi-channel outreach automation.
• Jasper. AI marketing copy and brand-voice content at scale.
• Blotato. AI social media content creation and repurposing.
• ElevenLabs. Best-in-class AI voice generation and audio production.
• HeyGen. AI avatar video creation for training, marketing, and personalised outreach.
• Obsidian. Local-first knowledge base for owning your notes and context outside any AI platform.
• Otter.ai. Meeting transcription and live collaboration notes.

SECTION 7: BUSINESS STARTUP

A complete startup curriculum, applied through your AI partnership.

For founders and aspiring founders, the portal includes a dedicated Business Startup section: eighteen guides taking you from first idea to operating business, each applied hands-on with Claude using the Bolt Method.

This document lists the curriculum so you can see the journey. It is an advanced optional training programme for AI partnership practitioners that have completed the Accelerator.

1. Founder Mindset & Identity
2. Start With Why
3. Market & Opportunity Research
4. Target Customer & ICP
5. Competitive Positioning
6. Business Model Design
7. Brand & Naming
8. Business Plan & Strategy
9. Legal & Structure
10. Financial Foundations
11. Funding & Investment Readiness
12. Building Your MVP
13. Sales Strategy & Pipeline
14. Marketing & Content Strategy
15. Network & Social Capital
16. Operations & Systems
17. Team & Hiring
18. AI Partnership for Startups

AI, DATA AND PRIVACY: PROBLEMS & SOLUTIONS

A practical guide from Bolt Practice. For SME business owners, investors and parents who want the upside without the exposure.

The One Idea to Hold

The Core Distinction. The risk is not "AI". The risk is what data you put where. Get that one distinction right and most of the exposure disappears.

What Follows. The rest of this guide is simply how to act on it; practical, clear, and sized for real people and SME businesses.

The Mistake Many People Make

The Common Belief: "Free is risky, paid is safe." It is not that simple.

The Reality. On consumer tiers like Claude Free/Pro/Max, ChatGPT Free/Plus, and Gemini Free/Advanced, your conversations can train future models, often on by default. You can turn this off in your settings, but £20 a month usually buys better features, not better privacy. The real line is a consumer account versus a commercial agreement with a Data Processing Agreement.

Training Off Is Not the Same as Compliant

Turn Off Training. Turning off the training setting stops your chats improving the model. You should do this on ALL consumer accounts.

But It Is Not Enough. It does not give you a processor agreement, defined retention, or audited access; the things that make you defensible as a data controller under GDPR.

What You Actually Need. A processor agreement, defined retention, and audited access come only with a commercial agreement.

The Rule of Thumb

• Never put client, customer, or confidential information into a consumer-tier account.
• Consumer tiers are for research, non-confidential drafting, and image generation. Nothing that names a real client or reveals confidential information.
• Real-world example. An estate agent uploading a client Sales Memo to a free chatbot breaches GDPR and confidentiality in one move. Wrong account for the data; not wrong to use AI.

Five Solutions, from Enterprise to On-Device

A ladder from highest governance to most private. You do not need the top level; you need the rungs that match your size and your most sensitive data.

1. Enterprise Platforms. Full governance for large organisations.
2. Team / Enterprise AI Accounts. The practical sweet spot for SMEs.
3. Self-Hosted Open Models. Data never leaves your infrastructure.
4. Anonymisation. A free habit for everyone, available immediately.
5. On-Device Offline (on a phone, tablet or computer). Maximum privacy, no internet required, nothing leaves your device.

Solution 1: Enterprise Platforms

Who It Is For. Large organisations with complex compliance requirements and dedicated IT teams.

What It Offers. Platforms like ServiceNow have solved governed AI for the world's biggest companies: customer data protected in a controlled environment, with a full audit trail and compliance frameworks.

The Honest Caveat. Out of reach for micro-businesses and most SMEs on cost and complexity. If this is not your rung, move to Solution 2.

Cost: Setting up the software requires certified consultants, with deployment costs routinely starting at £25,000+, even for basic setups.

▸ Watch the ServiceNow keynote (May 2026); "Control Tower" starts at 18m 45s

Solution 2: Team and Enterprise AI Accounts

Recommended for SMEs. The practical sweet spot for most small and medium-sized businesses.

• No training on your content. On a Team or Enterprise plan (e.g. Claude), the provider will not train on your content.
• Data rights and DPA. Your organisation keeps rights to its data; you get admin controls, audited access, and a Data Processing Agreement.
• Modest cost. Cost is per seat, but modest against the risk it removes.

Start: upgrade to a Team plan, appoint one owner, and write, publish and share a policy document.

Example: Claude Team Plan; Enterprise-Grade Security

No Model Training on Your Data

• Complete IP protection: your data, prompts, and file uploads are strictly confidential.
• Privacy assurance: Anthropic never trains its generative models on your company's interactions.

Identity & Access Management

• Domain capture: automatically blocks unauthorised personal accounts using company emails.
• Enterprise SSO: integrates seamlessly with Okta, Google Workspace, and Microsoft Entra ID.
• Smart provisioning: Just-in-Time (JIT) onboarding creates secure seats instantly for new hires.

Workspace & Data Governance

• Isolated Projects: internal documents and shared knowledge bases remain private by default.
• Role-based controls: admins manage exactly who can view, edit, or share specific information.
• Connector guardrails: global administrative control over integrations like Google Drive and GitHub.

Visibility & Compliance

• Audit logging: comprehensive tracking of user access, data footprints, and activity.
• MCP restrictions: strict administrative allow-lists for secure terminal-based tools like Claude Code.

Claude Team Plan: Pricing & Tiers

Flexible plans built for business collaboration. Pricing accurate as of 15 July 2026; check claude.ai for current pricing.

• 5-seat minimum: all Team plans require a baseline commitment of at least 5 users.
• Data privacy included: enterprise-grade security and no data training on all seats.
• Mix and match: blend Standard and Premium seats seamlessly on one dashboard.

Budget estimates (5-user minimum): Standard baseline starts at $125/month (annual) or $150/month (monthly). Premium baseline starts at $500/month (annual) or $625+/month (monthly).

Solution 3: Self-Hosted Open Models

Who It Is For. Regulated or high-sensitivity businesses (legal, healthcare, finance) where data must never leave your infrastructure.

How It Works. Run an open-weight model (Llama, DeepSeek, Mistral, Gemma) on your own servers. Data never leaves. This only works with open models; Claude and ChatGPT cannot be run privately, by design.

Cost Range. £0 to £1,500 for everyday quality (often hardware you already own, e.g. a laptop or desktop); £4,000 to £6,000 for near-frontier performance.

The Trade-Off. You buy control over your data, not peak performance.

Solution 4: Anonymisation

Free · Available Today · No Tools Required

A free technique for everyone; a habit, not a tool. When thinking through a real scenario, strip identifying detail before you paste.

Test before you paste: "Could a stranger identify the people or the deal?" If yes, use labels.

Solution 5: On-Device, Offline Models

For your most sensitive personal thinking; nothing sent anywhere, ever.

How It Works. Models like Google's Gemma run entirely on your phone; no internet connection, nothing transmitted to any server.

Getting Started. Free via the AI Edge Gallery app (iPhone App Store and Google Play Store). Up and running in about five minutes. Note: requires roughly 3 to 4GB of storage.

The Feature, the Limits, and the Workaround. Nothing is saved. Close the chat and it is gone; like a private notepad that shreds itself. Not a working long-term solution, but you can manually copy and paste the chat or a summary into a Google or Word document, then use find and replace to swap in anonymised labels (e.g. find and replace "Fred & Sally Bloggs" with "Client A"). It is a slow and cumbersome solution, but it is also free.

The Shadow-AI Problem

Banning AI does not work. People use it anyway, and a ban removes your visibility of the risk, which is often worse.

• [ ]% of knowledge workers use AI without organisational approval (Okta, 03/26)
• Executives believe they have visibility, yet over half of workers are using unapproved tools; a dangerous gap
• [ ]% of those using unapproved tools rely on free versions (BlackFog)

The fix: a sanctioned, governed, easy option so good nobody goes rogue.

Age, Young People and Oversight

LLM AI platforms do not share one minimum age. Know the rules before handing over a device. Accurate as of 15 July 2026; check each platform's terms for the current position.

Best Practice for Young People

Treat AI literacy like road safety: you teach it, you do not just ban it.

• Under 13. Full supervision. Parent's account.
• Ages 13 to 15. Supervised. Parent's account. Young person drives; adult reads responses.
• Ages 16 to 17. More independence on approved tools. Controls on. Periodic check-ins.
• 18+. Full access to appropriate platforms with proper account setup and policy awareness.

The Honest Limit

• No tool is magic. Controls get circumvented, rules get ignored, offline models are less powerful. No policy is bulletproof.
• The point is not perfection. It is matching protection to sensitivity, deliberately. A conscious, proportionate choice; not a perfect one.
• Scope it deliberately. Just because you can connect AI to everything does not mean you should. Scope it, permission it, keep a human in the loop.

Your One-Page Checklist

1. Pick the right account for the data: consumer for non-confidential, Team/Enterprise with a DPA for client-facing work.
2. Never paste confidential data into a consumer account.
3. Turn off training on consumer accounts; but know it is not compliance on its own.
4. Anonymise by default: Company A, Client B, Project Y.
5. Approve a tool rather than banning AI outright.
6. Keep an offline model for your most sensitive personal thinking.
7. Scope agentic access; keep a human in the loop.
8. Set age rules: 18+ for Claude; 13+ with consent for others; supervise the young.

Better Thinking. Better Decisions. Better Outcomes.

Bolt Practice and The Bolt Method™ help people and businesses think better with AI. Human-first, human-last, always.

• Never too frightened. Understand the tools well enough to use them with confidence; not paralysed by fear.
• Never reckless. Data stays protected. The right account for the right data, every time.
• Judgement stays human. Operate where judgement stays human, data stays protected, and the upside benefits everyone.

COMMON AI CONCERNS, ANSWERED

The questions you will be asked by others, and the questions worth asking yourself if you haven't already.

As you build your AI partnership, some people close to you will raise concerns; many are legitimate, and the Method's own Test for Truth demands they get honest answers, not dismissals. Here are the thirteen most common, with our answers.

1. "People will talk to AI instead of humans; won't that make us isolated?"

The Bolt Method is human-first by design: the human comes first and last, and the partnership exists to make you more present and capable with the people in your life, not less. At Network level it is explicitly shared, across teams and families. And honestly: billions of people suffer from loneliness. If AI can bring calm and clarity to someone who is alone and either it doesn't have a friend or family member they can talk to, or the money to pay for a coach, therapist or other qualified professional to support them, then, in most cases, something is better than nothing.

1. "Will my data be tracked, stored, or leaked?"

A legitimate concern with a practical answer: the risk is not "AI", it is what data you put where. The AI, Data and Privacy section of this guide gives you the full discipline; the right account for the right data, every time. You already trust your phone more than you realise; this makes that trust conscious.

1. "Won't this make me mentally lazy or creatively blocked?"

Used at Tools level, it can; that is exactly why the Method exists. The Judgement element keeps your thinking active (Test for Truth; your judgement stays accountable), and the partnership is built to challenge you, not to think for you. You are the driver; AI is the car. And overwhelm kills more creativity than any tool ever has; a good partnership opens the space to explore again.

1. "Is this part of the wave of tech that puts people out of work?"

The wave is real and pretending otherwise would fail our own Test for Truth. The Method trains you to partner with AI rather than compete with it; that is the difference between roles that disappear and roles that evolve. You cannot stop the wave, but you can learn to surf.

1. "What if people use it to deceive or harm others?"

They will; AI can be used to lie, and so can humans. It is about who wields it, and why. That is precisely why the Values element is embedded into your AI before anything else: honesty, care, restraint, stewardship, operating on every turn. And Test for Truth is your defence on the receiving end.

1. "What if I get too emotionally attached or dependent?"

The Method builds in the counterweights: rest, AI-free zones, Rule 5 of the practice (expect blocks; take breaks), and the Human level of collaboration, which never stops mattering. Millions are addicted to their phones and to feeds engineered for compulsion; a reflective conversation is the opposite of that pattern.

1. "If it's this good, won't it replace real coaches, therapists, and mentors?"

It amplifies them. Like a mirror helps a dancer, an AI partner helps coaches and therapists do better work and supports their clients between sessions. For anything regulated or clinical, the Method's position is explicit: set direction with AI but point to qualified human support.

1. "I'm not technical. I don't know how to use AI properly."

If you can talk to a friend, you can do this. The Method is voice-first, the setup in this guide is copy-and-paste, and whenever you are stuck, you ask the AI itself to walk you through it. No technical skill required; that is the point of this programme.

1. "What if I misuse it or lean on it too much?"

The Weekly Review exists for exactly this: a built-in check on how you are using the partnership, every week. Balanced use is a habit like any other. Think of it as cruise control; you are not giving up the wheel.

1. "Isn't this just another trend that will fade in six months?"

Tools and apps come and go. This is not a tool; it is an operating system for how you think, decide, and act, and it is platform agnostic by design; if today's platforms vanish, the Method moves with you. Trends fade. Operating systems become the way we live.

1. "I'm too busy. I don't have time for this."

Ten minutes a day is the entry point, and the time it returns dwarfs the time it takes; that is not a slogan, it is what the Bolt ROI Calculator measures. Pausing to tie your laces before a race is never wasted time.

1. "Isn't AI bad for the planet?"

The energy footprint of AI at industry scale is a real and live issue, and it deserves honest attention rather than a wave of the hand. At the individual level, your usage choices matter less than the industry's, but the stewardship principle still applies: use it purposefully, not compulsively. AI's impact depends on its intent; light touch, deep change. At Bolt Practice, we have an initiative led by Ella Bolton called EarthStream. Think of it like recycling for the environment. It’s a user-level initiative that can enable individuals to reduce the environmental impact of their AI use.

1. "We don't know how Big Tech trains these systems; isn't bias baked in and hidden?"

Partly, yes; that is why AI Bias is one of the four positions on the Bias Matrix, with its own defences (Engage Myriad Minds; cross-verify across models and sources). It is also why platform choice is a values decision: we recommend the provider whose approach to transparency and safety we judge best aligned. AI does not have to reflect the worst of the internet; with transparency and ethical oversight, it can reflect our better values instead.

THE BOLT PATHWAY: FROM DISCOVERY TO PRACTITIONER AND BEYOND

A clear path from first step to full practice. Each step moves a person or business further up the Five Levels of Collaboration.

AI access is becoming normal. AI capability is not. Most people never move beyond Level 2; the market is selling Tools and Agents as the destination. Bolt Practice teaches people to reach Partnership and Network. The pathway below is how; each rung builds on the one before it, and this guide is your companion for the second rung.

Beyond the ladder, Bolt Practice runs RiZE (the free youth and Gen Z employment programme) and, in development, Educating Alpha (AI tutoring for under-18s, parent-managed); social impact offers that sit outside the pathway but carry the same Method.

The AI Partnership Accelerator

Your Human-AI Partnership Transformation. Better thinking, better decisions, better outcomes.

Most people use AI as a faster way to do the same old things.

A few people use it to think, decide, and operate at an entirely different level. The gap between the two is not technical skill; it is a method and the Accelerator shows you how.

This is not a course you finish. It is a daily practice you keep for life.

What changes for you

The Accelerator builds your AI capability and your confidence first; everything else follows.

• Save Time. The work that ate your week starts taking hours.
• Save & Make Money. Better thinking shows up on the spreadsheet.
• Think More Clearly. A structured way to move from vague intent to a sound decision, every time. The benefit underneath all the others.
• Build a Capability That Stays. You leave with habits and a working AI Partnership; not a folder of prompts you'll never open again.

Real People, Real Change

Caroline Marsh, Care Home Owner. Two-day creative tasks now take ten minutes.

"I'm not sure where I would be if it hadn't been for the last 12 weeks. The impact it's had on my life, my family and the business has been incredible."

Neil Mansell, Property Investor. A £400,000 decision resolved in hours, not weeks.

"I wasn't lacking desire. I was lacking a system. AI is now part of how I think, decide and operate every day."

Phil Geraghty, Co-founder, Crowdfunder.co.uk. Thousands of users supported daily for under £1 a day.

"The walk and talk habit changed everything for me. Voice-first AI Partnership unlocked thinking I didn't know I was carrying."

97% of participants rated the Bolt Practice Accelerator as transformational, with an average of 60 hours of time saved on average for those that completed the programme.

How the 90 Days Work

The Accelerator runs as a 90-day programme built around the Three Core Components: you (the human in the loop), The Bolt Method, and AI capability. Every element of the programme develops all three in parallel.

• Fortnightly group coaching calls. Six live calls across the 12 weeks, all recorded, with slides and videos shared in the community.
• Fortnightly live drop-in sessions between calls. Live Q&A and troubleshooting with coaches; bring real work and real questions, get hands-on guidance, and connect with fellow practitioners.
• Circle community and content. Your space to connect, ask, and share between calls.

The Four Critical Success Factors. For the next 90 days, and for any serious effort beyond it:

1. Commitment and motivation. You have to want this. Pain changes behaviour; pleasure sustains it.
2. Mentors and coaches, human and AI. Form a council of perspectives you respect, and use them.
3. Community. You become the average of who you spend time with. Surround yourself with people further along, not further behind.
4. Ongoing practice. This is a practice, not a one-off course. Professionals keep practising for their entire career; so should you.

The Six-Call Programme

Each call develops all Three Core Components side by side.

Measuring Your Return: The Bolt ROI Calculator

Most businesses talk about AI value in round numbers nobody can check. Bolt Practice built a method that survives scrutiny, for everyone who uses it, not just the person who owns the business. The Bolt ROI Calculator rests on three disciplines:

• Every AI conversation counted, the ordinary ones alongside the impressive ones; never a highlight reel.
• Work achieved valued at real market rates, never inflated, with two honest lenses, what the work would have cost and what your own time is worth, that are never added together.
• Every error the AI catches before it costs something, logged as its own proof, separate from the value of the work itself.

In the internal pilot (one founder, one month, four ventures): 786 hours of advisor and staff-equivalent work achieved, £150,575 in advisor or staff equivalent cost, and 18 risks and errors caught and avoided; among them a maths error on a live fundraise and a personal-liability warranty risk. The honest number is the sum of what was harvested, never a projection. And if you remain sceptical, discount it by 90%; that is still more than 78 hours and £15,000 of equivalent work in a single month.

During the Accelerator you will run a simplified version on your own Claude usage, so you finish the programme with your own measured number, not just our word for it; it becomes your ROI presentation in the final call. The calculator is in beta testing now.

The Bolt Method™ is a trademark of Athena Veritas Ltd; trademark application filed, registration pending. Formerly published as The Veritas Method®, a registered trademark of Athena Veritas Ltd.

Bolt Practice is a trading name of Athena Veritas Ltd.

© Athena Veritas Ltd 2026. This guide is provided for programme participants' personal use. Please do not copy, share, or teach from it without written permission.