Case 01 / LifeOS

A source-of-truth architecture for AI-assisted life and work

LifeOS is a private, portable context system that separates current state, durable memory, raw evidence, decisions, operating rules, and repair loops so AI-assisted work stays grounded.

Why

Context that can survive beyond a chat.

Useful AI work breaks when the source of truth lives across chats, notes, screenshots, documents, apps, and human memory. The problem is not just recall. It is routing: what is current, what is durable, what needs evidence, and what should change when the system is wrong.

LifeOS started as an answer to that fragmentation. Not another notes app. Not a productivity dashboard. A private, file-based source of truth where context can stay portable, inspectable, and useful for both human decisions and AI-assisted work.

Before

Context kept restarting

Useful AI work still depended on re-explaining the same background across chats, projects, files, and tools.

Current state and durable memory mixed together

Temporary operating context, stable facts, raw evidence, and old conclusions needed clearer boundaries.

Corrections disappeared

A bad answer could be corrected in the moment without changing the system that produced the failure.

Evidence had no reliable route

Screenshots, records, notes, logs, decisions, and source files needed owners instead of becoming vague memory.

Operating model

Source files Markdown domains, context files, indexes, logs, and raw evidence

Rules AGENTS.md, source hierarchy, routing, and update boundaries

Reasoning ChatGPT reads context and helps decide what the situation means

Operations Codex inspects files, patches source, checks diffs, and verifies state

Repair Repeated failures become small changes to the source of truth

How it works

01 Capture Important context, evidence, decisions, and current state enter the system.

02 Route Each item goes to a domain, current layer, durable memory, or raw evidence.

03 Work Human and AI reason from the same maintained source instead of loose memory.

04 Update When reality changes, the source files change instead of staying in chat.

05 Repair Repeated failures become small fixes to routing, context, or source hierarchy.

Example

A correction becomes a system patch.

The proof is not that AI gets every answer right. The proof is that a recurring failure can be traced, sanitized, and converted into a small source-of-truth repair.

Failure

AI answered from stale or incomplete context. The useful signal was an unclear source boundary.

Diagnosis

The issue was not just the model. The routing rule, source file, or update path was unclear.

Patch

Update the relevant file, routing rule, or context boundary so the correction becomes part of the system.

Future behavior

The next similar request is grounded in the corrected source instead of relying on chat memory.

Built

markdown source-of-truth domains

domain-level routing rules

NOW / MEMORY boundaries

raw evidence separation

Codex file-operation loop

stale-context checks

compact update reports

failure-to-repair workflow

What changed

The useful part is not more memory. It is maintained understanding.

LifeOS does not try to remember everything. It decides where information belongs, what needs evidence, what is temporary, what is durable, and when the system should be repaired instead of simply corrected in chat.

AI can reason against maintained files instead of loose chat memory.
Current context, durable facts, raw evidence, and operating rules have separate places.
Repeated workflow failures can become small source-of-truth patches.
The system stays portable: files remain readable, editable, inspectable, and recoverable.

What this points to

LifeOS is the large version.
Client work starts smaller.

This is why the entry point is an AI Workflow Diagnostic. It starts with one recurring workflow where AI keeps losing context, evidence, decisions, or review logic.

The diagnostic is not a whole operating system build. It is a focused way to find the source-of-truth gap, context boundary, or repair loop that should be fixed first.

Want to see the seven domains?

The Journal note shows how LifeOS is split across System, Work, Health, Nutrition, Lifestyle, Public Work, and Finance.

Read the seven-domain note