Introduction

View as Markdown

What is XTrace?

XTrace is a belief revision system for AI. It gives you persistent context across ChatGPT, Claude, Gemini, and every AI tool you use — not by dumping chat logs into a vector database, but by maintaining a structured model of what’s true, revising it as things change, and strengthening it with every correction.

The storage problem is solved. The maintenance problem isn’t. Memory without revision degrades into noise. XTrace solves maintenance.

Why beliefs, not memories

Every AI memory product stores facts. But facts go stale: you change jobs, drop a hobby, reverse a decision, pass a deadline. The facts you stored are now wrong, and the system has no idea.

XTrace maintains beliefs — assertions about the world that can be revised, retracted, expired, and protected based on how much they matter.

MechanismWhat it does
EntrenchmentBeliefs carry weight. What you explicitly state outranks system inferences. Core beliefs are protected from silent overrides.
ContractionBeliefs can be removed cleanly — “we’re NOT doing the Kafka migration” — without needing a replacement. The system distinguishes real-world changes from its own mistakes.
Dependency propagationWhen a belief changes, everything built on it is flagged — downstream beliefs and artifacts. No hidden breakage.
Learning from correctionsEvery correction becomes a labeled example. The system retrieves similar past mistakes before future extractions, improving over time without fine-tuning.

Three types of knowledge

XTrace extracts and maintains three kinds of structured knowledge from your conversations:

Beliefs (facts)

Atomic assertions: “user prefers TypeScript,” “launch window is 30 days post-GA.” Each carries entrenchment weight, can be superseded or retracted, and links to the conversation it came from.

Episodes

Conversation summaries that capture the arc of a session — what was discussed, what was decided, what’s still open. Episodes link to the beliefs they established.

Artifacts

Work products: a technical design, a blog post, a GTM strategy. Version-chained like git — each revision links to the previous, with the beliefs and decisions that drove the change.

No more cold starts

Every AI session today is a cold start. You re-explain yourself, re-establish context, re-upload your work.

XTrace eliminates this. At the end of every session, the system extracts what matters: the beliefs established, the decisions made, the artifacts created or revised, the open threads. When you start a new session — same tool, different tool, next day, next month — the context is already there.

Session 47 feels like session 1, but with full context.

Cross-tool continuity

Draft a strategy in Claude. Research competitors in ChatGPT. Build the slide deck in your favorite tool. Each connects to the same belief system. Each starts with the same context. Each tool’s contributions flow back.

Corrections propagate too. Tell Claude “we’re NOT doing the Kafka migration” and every connected agent updates. One correction, everywhere, immediately.

Context as login

The endgame: “Connect your XTrace” gives an app your preferences, your work history, your decisions — scoped to exactly what that app needs. You don’t onboard tools anymore. Tools onboard to you.

What’s next