Core Concepts
What is a yawn.bot?
How the six dimensions of inquiry turn rough ideas into structured, executable plans. The public mirror lives at /yawnbot.
What is a yawn.bot?
A yawn.bot is the working interface to Yawn — a specification engine that turns rough ideas into structured, executable plans. Where a chatbot gives you answers, a yawn.bot gives you clarity. It does this by asking questions, propagating constraints, surfacing tensions, and building confidence until your intention is sharp enough to act on.
A yawn is a living relationship contract between you and something you care about. It captures your raw intention, extracts what matters, structures it into goals and constraints, then runs experiments to close the gap between where you are and where you want to be.
Part journal, part scientific method, part AI partner — a yawn evolves with you through structured dialogue, evidence, and continuous learning. It is not a task list. It is a coherence system for your intentions.
The six operators
Every yawn has six operators that work together to help it understand what's missing and what to do next. Think of them as sections of its brain. Each one does a different kind of work. Together, they turn confusion into understanding, and understanding into action.
| # | Operator | What it does | |---|----------|-------------| | 1 | The Scout | Figures out what is going on. Finds the situation, the context, and where to begin. | | 2 | The Gap Finder | Looks for what is missing. Spots the unclear part, the blind spot, or the unanswered piece. | | 3 | The Reframer | Checks whether you are looking at the problem the right way. Sometimes the fastest progress comes from seeing the same thing differently. | | 4 | The Priority Keeper | Figures out what matters most. Separates signal from noise and brings the priority into focus. | | 5 | The Committer | Turns insight into a real next move. Helps decide what you are willing to do, test, or choose. | | 6 | The Overseer | Steps back and looks at the whole picture. Notices patterns, connects things, and decides how the process itself should change. |
No single operator solves the whole problem. One locates the issue. One finds the missing piece. One changes the frame. One finds what matters. One turns it into action. One watches the whole process. Together, they form a whole yawn.
The six dimensions in detail
Every yawn.bot conversation is shaped by these six operators working behind the scenes. They aren't separate bots — they're dimensions of inquiry that the system applies to whatever you bring it.
1. Orienting — "What's going on here?"
The bot starts by scanning what you've given it — a messy note, a half-baked idea, a link, a frustration — and building a map of the situation. It extracts entities (people, projects, deadlines, amounts), identifies the domain (financial, professional, personal, creative), and connects your input to anything you've already been working on.
This is literal pattern recognition: what are the nouns, what are the relationships, where does this sit in the context of everything else you care about.
2. Gap-finding — "What's missing?"
Once the situation is mapped, the bot looks for what isn't there. Missing constraints. Unstated assumptions. Dependencies you haven't named. Questions you haven't asked yourself yet.
This matters because most plans fail not from what's in them, but from what's absent. The bot's job here is to surface the blind spots — the things you'd discover the hard way three weeks into execution.
3. Reframing — "Is the question even right?"
Sometimes the most useful thing isn't answering your question — it's questioning your question. "I need to find a new job" might actually be a meaning problem, a courage problem, or a financial runway problem. Each reframe opens a different path.
The bot checks whether the frame you're using is the most productive one, and offers alternatives when a reframe would unlock better options.
4. Implication-tracing — "What follows from this?"
Most tools work in the present tense. Yawn thinks in futures. The bot takes your constraints and choices and traces their consequences forward. If you set this budget, what can't you afford? If you choose this architecture, what maintenance burden do you inherit? If you keep doing this for a year, where do you end up?
This is constraint propagation: one answer changes other answers, and the bot shows you the chain reaction.
5. Commitment — "What will you do about it?"
Insight without action is trivia. The bot turns analysis into testable experiments with concrete timescales, success criteria, and branching logic (what to do if it works, what to do if it doesn't).
Not "you should consider..." — but "here is a 14-day experiment that would prove or disprove this hypothesis." The system treats your goals like science: hypotheses, evidence, falsifiable claims.
6. Meta-oversight — "Is this inquiry on track?"
The final dimension looks at the inquiry itself. Is the yawn zoomed in too far? Should it decompose into sub-yawns? Has confidence stalled? Is the bot asking the right questions, or is it stuck in a loop?
This is the quality-control layer that prevents the system from going through the motions without making progress.
How it works
You bring in something unresolved. Yawn looks for what is missing. The operators work together to understand it. You answer, correct, or redirect. Together, you turn confusion into a clearer next move.
Step by step
- Bring in something real — Say what is on your mind, even if it is messy.
- Let Yawn find the gap — Yawn starts looking for what is unclear, missing, or important.
- Respond to what shows up — Confirm it, correct it, or choose a different direction.
- Follow the next move — Keep exploring, take action, connect it to another yawn, or save it for later.
- Return with proof — What changed? What did you try? What did you learn?
How you win
You win when the situation is clearer than before and you know what to do next. Sometimes the win is an answer. Sometimes the win is a better question. Both count.
The rule: Understand enough to move. Move enough to understand better.
How confidence works
The bot doesn't use a single confidence number. It tracks four dimensions:
| Dimension | What it measures | |---|---| | Coverage | What percentage of required variables have been resolved? | | Consistency | Are there contradictions or unresolved dependencies? | | Sufficiency | Is there enough information to choose a next move? | | Proof readiness | Are the success criteria specific and testable? |
These four scores combine into a composite confidence that determines how much autonomy the system has earned. Low confidence means the bot asks more and assumes less. High confidence means it can suggest answers, auto-fill derived values, and propose action — but only where the evidence supports it.
What a .yawn file contains
When you work with a yawn.bot, the structured output is a .yawn artifact. It holds:
- Intent: what you're trying to accomplish, in your words
- Constraints: hard limits (budget, deadlines, regulations) and soft preferences (style, risk tolerance)
- Tensions: conflicts the system has detected between competing goals or constraints
- Confidence vector: the four-dimensional score described above
- Experiments: testable hypotheses with success criteria and branching logic
- Evidence: proof collected from experiment runs
- Next move: the single most valuable action given current confidence
A yawn isn't a document you write and file away. It's a living specification that updates as you answer questions, run experiments, and collect evidence.
How it differs from a chatbot
A chatbot optimizes for a response. A yawn.bot optimizes for resolution.
| | Chatbot | Yawn.bot | |---|---|---| | Goal | Produce an answer | Narrow ambiguity until action is safe | | Memory | Conversation history | Structured specification with constraint propagation | | Questions | Clarifying follow-ups | Scored by expected information gain | | Output | Text response | Living .yawn artifact with confidence, tensions, experiments | | Trust | Implicit (you trust the output or you don't) | Explicit trust gates based on measured confidence | | Failure mode | Confident-sounding wrong answer | Transparent uncertainty with a plan to resolve it |
The holarchy: yawns that contain yawns
When a problem gets complex enough, a single yawn decomposes into child yawns — each focused on one piece of the puzzle, each inheriting constraints from the parent. The parent tracks the children's progress and propagates their results back up.
This is how the system handles real complexity without losing coherence. A startup pitch might decompose into market validation, financial model, product spec, and hiring plan — each its own yawn, each contributing proof back to the parent's confidence score.
Who it's for
Yawn.bot is built for anyone who knows roughly what they want but doesn't yet have it in a form that can be safely executed — by themselves, by a team, or by AI agents.
Founders turning ideas into ventures. Product managers turning requirements into specs. Individuals turning life goals into testable plans. Teams turning strategy into coordinated action.
The common thread: you have intention but not yet clarity, and you want a system that helps you earn that clarity through structured inquiry rather than guesswork.
FAQ
Why not just prompt an AI?
Prompting skips the hard part: getting clear on what the real goal is, what constraints matter, what is still unknown, and what proof would count as success. A prompt gets you output. A yawn gets you a specification you can trust.
What are trust gates?
Trust gates determine how much autonomy the system has earned at any point. Early on, the bot mostly asks. As confidence increases across all four dimensions, it can suggest answers, auto-fill inferred values, and eventually support near-autonomous execution — but only where the evidence justifies it.
Is a yawn.bot replacing human judgment?
No. It's a tool for making human judgment more precise. Yawn handles what can be inferred and computed. Humans stay in the loop where authority, values, or irreducible judgment calls are required.
How does the bot know what to ask next?
Questions are ranked by expected information gain — how much a given answer would narrow the remaining uncertainty. The goal isn't to ask more questions; it's to ask better ones.
Can I use this for anything, or just tech/startup stuff?
Anything with intention, constraints, and uncertainty. Career decisions, creative projects, personal goals, business strategy, product specs, research questions. The six dimensions of inquiry apply everywhere.
What does "specification engine" mean?
It means Yawn's job is to produce a spec — a structured, precise, testable description of what you want and how to get there. Not a document, but a living artifact that evolves through inquiry and evidence.