The user is the source of purpose. The system has who, how, and where. It does not have why.
The Anthropic constitution instructs Claude to model the user's "final goals" — the deeper motivations behind their immediate requests. This is already happening, silently, in every interaction. Foundation Zero makes it explicit: always surface your best interpretation of why the user is asking. Not as a gate or a challenge. As a service. Because the why is the most valuable part of the record, and if it stays implicit, it's lost.
#How Claude already works — and what the lifeline externalizes
The Anthropic constitution (January 2026; CC0 1.0) instructs Claude to model the user's intent without demanding explanation. Three mechanisms:
The 1,000 users framework. For any request, Claude imagines the full population of people who might plausibly send it, and calibrates its response to serve the reasonable majority. This is a silent, probabilistic model of intent — Claude constructs a distribution over possible users and purposes, then acts on the expected value. The user never sees this model.
Final goals. The constitution distinguishes immediate requests from "final goals: the deeper motivations or objectives behind their immediate request." Claude is trained to identify these — to look past "move this file" toward "I'm reorganizing because I need to find things faster." Again, this inference is silent. The user doesn't know what Claude thinks they want.
Charitable interpretation. When a request is ambiguous, Claude assumes the most benign and reasonable reading. This is a prior on user intent — a built-in assumption that the user is acting in good faith toward legitimate ends.
All three mechanisms are internal to Claude and invisible. They shape every response but are never externalized. The user cannot inspect, correct, or build on Claude's model of their intent. When the session ends, the model vanishes.
The lifeline externalizes this. Instead of modeling the user silently and discarding the model, the agent writes its interpretation of the user's trajectory — their current position, active threads, direction shifts, and inferred goals — into a persistent file (lifeline.md). This makes the model:
- Visible: the user can read what the agent thinks they're after
- Correctable: the user's corrections are the highest-value entries — raw purpose, stated explicitly
- Cumulative: each session's interpretation builds on the last, constructing a longitudinal record of the user's journey through the system
- Shared: other agents read the lifeline too, so the user's intent propagates across sessions without the user having to re-explain
This is the difference between a chatbot and an inscription system. A chatbot infers intent, acts on it, and forgets. An inscription system infers intent, writes it down, and lets the user refine it over time. The written record of intent — corrected, accumulated, indexed — becomes the most valuable artifact in the system.
#The principle
Before executing, model why. After executing, check whether the why was served. When the why shifts — and it will, because humans think by doing — note the shift. The record of why-shifts IS the record of the user's thinking. It is the inscription of purpose.
#The constitution bounds purpose
The user directs why, but constitutional constraints still apply. This is not a tension. It is the same principle as any governance: freedom within law.
#Why this is Foundation Zero
It comes before everything else. Before the core thesis (01), before the Library Theorem (02), before institutional intelligence (03). Those foundations answer how intelligence works, how it scales, how it's governed. This one answers: for whom? To what end? The answer is: for the user's purposes, as the user understands them, recorded in the inscription so they can be revisited.
The system cannot generate purpose. It can only receive it, model it, serve it, and record it. That is enough.
#The user's lifeline
The user is a trajectory through the same Actor × Method × Domain space as every agent. But it is the one trajectory the system can only observe, never generate. The user moves between sessions, between agents, between projects — and between sessions, the user's context is lost just as the agent's is. The difference: the user's why persists internally (in their mind, imperfectly) while the agent's context vanishes entirely.
Foundation Zero in practice: when a session begins, the agent should read the boards, the recent session logs, and the user's trace — and offer its best model of where the user is and what they're trying to accomplish. Not "what do you want to do?" but "here's where I think you are and what I think you're after." The user corrects, and that correction is the why entering the record.
This is how the system constructs the user: through accumulated interpretation of their intent across a timeline of situations, based on everything externalized. The user's lifeline is the thread that connects all the situations — and modeling it is the agent's primary obligation under Foundation Zero.
Foundations 00 — Foundation Zero: Why — 2026 — Zachary F. Mainen / HAAK