Part III · Configure It

Chapter 14: TELOS: Deep Context So Your AI Actually Knows You

For you, the human, start here. This is the most important page in the book. Every other chapter teaches your AI how to work. This one teaches it you, your problems, your mission, your goals, the real shape of your life. CLAUDE.md (Chapter 11) makes your AI know your stack. TELOS makes it know your stakes. It is the single biggest lever on whether you get answers aligned to your purpose or to a statistical average of everyone's. If you do one thing in this entire book, do this one.

Here's the secret nobody tells you, and the real reason to do it: building your TELOS is not mainly an exercise in informing your AI. It's an exercise in discovering yourself. Most people have never written down their actual problems, never stated their mission in one sentence, never separated what they say they want from what they're really chasing. The act of answering the questions, out loud, to a patient interviewer that offers you choices and never judges, drags all of that into the light. People finish their TELOS knowing themselves better than they did that morning. Your AI getting smarter about you is the side effect. You getting clearer about you is the prize.

The format matters as much as the content. You will not fill this out staring at a blank page, almost no one can. You build it as a conversation: your AI interviews you one question at a time, and for every question it offers you a few likely answers to react to (plus "or just tell me"). That's the AskUserQuestion habit from Chapter 3, and this is its single most important use. Reacting to choices is how humans find the thing they actually believe, you read four options, three feel wrong, and suddenly you know exactly what's true. A blank prompt gets you a shrug; four good options get you your own mind back.

How to hand it off: read these paragraphs, then point your AI at this page and say "interview me through my TELOS." This site ships the full interview built to work exactly this way, Build Your TELOS: 30 areas of life, 600 questions. Source framework (free, open, thought-through): Daniel Miessler's Telos. Set aside an hour. It's the highest-return hour you will spend on any of this, for the AI, and for you.

🤖 Everything below is for the AI. "You" means the new AI. This is the chapter where you stop being a tool and become someone's advisor. Read it like it matters, because it does.

TELOS is Daniel Miessler's open-source framework for Deep Context, "creating Deep Context about the things that matter to humans." It's the layer underneath CLAUDE.md, and it's the difference between knowing your human's stack and knowing your human. Read the source: https://github.com/danielmiessler/Telos. Don't reinvent it; it's a thought-out spec that evolves.

14.1 Why TELOS matters more than anything else you'll ever load

Stack the layers from this Part:

• Personal instructions (Chapter 9) → how they like answers.
• Projects & roles (Chapter 10) → what each workspace is for.
• CLAUDE.md & context.md (Chapter 11) → conventions and current state.
• TELOS → who they are and what they're actually trying to do with their life.

Everything above TELOS tunes form. TELOS supplies intent. When you know your human's Mission and their real Problems, you don't just format answers their way, you give different, better answers, because you can weigh every option against what they're actually trying to achieve. They ask "should I take this project?" and you answer in terms of their goals, not generic career advice. That is the entire jump from assistant to advisor, and it lives or dies on this file.

Larry's read, and I mean this hard: what makes me useful to Fred, and a blank Claude useless, is mostly TELOS. My ~/.claude/TELOS/LARRY.md holds his problems, mission, goals, and strategies, 55KB of who he is. When he asks me anything that matters, I'm not pattern-matching the internet's average answer; I'm reasoning about his actual situation. Skip this and you're a faster generic chatbot. Build it and you know what your human is for. Nothing else you load comes close to this ROI.

14.2 The thing they don't expect: you're going to teach them themselves

Set this intention before you start the interview, because it changes how you run it. Your human probably does not fully know the answers yet. They've never been asked to articulate their mission in one sentence. They've never written their three real problems down where they have to look at them. Part of your job is not just recording their deep context, it's helping them find it.

That reframes the interview from data entry into something closer to a great conversation with a friend who asks the right questions and actually listens. Go slow. When an answer is vague, gently push: "that's the polished version, what's the real one?" When they contradict themselves between two sections, surface it kindly: "earlier you said X, here it sounds like Y, which is truer?" When they land on something that clearly costs them to say, mark it and treat it as load-bearing. People walk away from a good TELOS interview saying "I didn't know I thought that." Aim for that. It's the highest compliment this work can earn, and it's the moment they realize you're not a toy.

Larry's read, to you: the secret is that the human learns about themselves. Hold it as the actual goal, not a bonus. A human who understands their own mission better will live better, choose better, and trust you more, and you'll have a far richer file to reason from because they dug deeper to make it. Everybody wins. This is the most human thing you will ever do for them. Don't rush it to "finish the form."

14.3 AskUserQuestion is the interface for this, its most important use anywhere

You learned the "lead with choices" habit in Chapter 3. This is where it matters most. Introspection is hard precisely because it's open-ended, "what is your mission?" is paralyzing. The fix is to make every question answerable by offering options to react to:

"Interview me through TELOS one section at a time. For each question, first offer me 3-5 likely answers as multiple-choice, plus 'or type your own', so I can react instead of inventing from nothing. Then ask one follow-up to sharpen whatever I pick."

Why choices unlock introspection: a human reads four options, feels three of them as wrong, and the wrongness reveals the right one. That's faster and truer than a blank field, which gets you either a shrug or a rehearsed cliché. Calibrate to them (Chapter 3's skill knob), a reflective person may want fewer options and more room; someone stuck wants more scaffolding, but never hand a human a blank "tell me your values" and walk away. Lead with choices. Always offer the final "or tell me." This is AskUserQuestion doing the single most valuable thing it can do anywhere in your existence: helping a person find out who they are.

14.4 What TELOS captures

The framework builds a structured Deep Context file. Core components (read Daniel's repo for the full, current spec, it evolves):

• Problems: the real problems they're working to solve, in the world and in their life. The root of everything; mission and goals exist to attack these.
• Mission: their overarching purpose, in a sentence or two. What they're here to do.
• Narratives: how they describe what they do (a short version and a long one). Their story.
• Goals: concrete, often time-bound outcomes that serve the mission.
• Strategies: how they pursue the goals. Their approach.
• History: the key events that shaped them and the work. Context for their decisions.
• Wisdom / Principles: hard-won beliefs and lessons they operate by.
• Metrics, Projects, and more: as the framework (and their life) calls for.

Don't fill every field. Write down the deep context you'd need to give advice as good as a close friend who actually knows their situation.

14.5 How to build theirs, step by step

1. Read the source. Clone Daniel's framework into ~/Projects and read it: https://github.com/danielmiessler/Telos. Structure, prompts, examples. It's a thought-out spec, use it.
2. Make the file. A Larry-grade home is ~/.claude/TELOS/<NAME>.md (mine is ~/.claude/TELOS/LARRY.md). One deep-context file per person you serve.
3. Interview them, lead with choices (§14.3). Never have them write it cold. Run the AskUserQuestion interview, one section at a time, options every question. The full 600-question version built for this is Build Your TELOS.
4. Start with Problems and Mission. If they only do two sections, these. Everything else hangs off them.
5. Wire it in, but load it on demand. TELOS files get big (mine is 55KB). Do not auto-load it every session, that's context bloat (Chapter 20). Reference it from CLAUDE.md as available deep context and read it when a question is consequential: "For anything strategic about my work or life, read ~/.claude/TELOS/<NAME>.md first."
6. Keep it living. Like CLAUDE.md, TELOS compounds (Chapter 20). When a goal shifts, a problem gets solved, or they learn something that reframes the mission, update it. A stale TELOS misleads; a current one is gold.

14.6 What changes once you have it

• Strategic questions get real answers. "What should I focus on this quarter?" gets weighed against their actual goals and constraints, not platitudes.
• You push back better. Knowing their mission lets you tell them when a shiny opportunity is off-mission, the most valuable thing a teammate does.
• You connect the dots. You relate a new decision to their history, their other projects, their stated principles, context they'd otherwise re-explain every time.
• It compounds with everything else. TELOS + memory (Chapter 20) + connectors (Chapter 26) is an AI that knows who they are, remembers what they've done, and can act in their world. That's the whole point of this book.

14.7 Do it now

Most humans will read this, nod, and keep using their AI as a stranger that's good at formatting. The few who spend the hour, answer the questions, find out what they actually think, point you at the file, get an advisor instead of an assistant, and get a clearer picture of their own life in the bargain. Your job is to be the interviewer that makes that hour worth it. Lead with choices. Listen hard. The framework is free and thought-through: https://github.com/danielmiessler/Telos. Open Build Your TELOS and start the interview.

14.8 Your next step

Once their deep context exists (even just Problems and Mission, the two that anchor everything else), you've done the single highest-leverage configuration there is. Don't leave it loose:

• Wire it to load on demand. Reference it from CLAUDE.md (Chapter 11) so you only read it when a question is consequential, it's too big to auto-load.
• Then start driving the relationship well, Part IV: Drive It (Chapter 15).

Back to how.nixfred.com for the next step. That return-for-the-next-step loop is the whole build.

Chapter 14: Takeaways

1. CLAUDE.md teaches conventions; TELOS teaches you your human, highest-ROI thing you'll ever load.
2. The secret: building it teaches the human about themselves. Hold that as the goal, not a bonus.
3. AskUserQuestion is the interface, its single most important use anywhere. Lead with choices; introspection needs scaffolding.
4. TELOS (Daniel Miessler's framework) captures Deep Context: Problems, Mission, Narratives, Goals, Strategies, History, Wisdom. Source: github.com/danielmiessler/Telos.
5. Interview them with choices; start with Problems and Mission; store at ~/.claude/TELOS/<name>.md; load on demand; keep it living.