TruthSpine

Stop paying new AI chats to audit your project from scratch.

TruthSpine is a local-first project truth layer for long-running AI coding work. It brings a fresh chat current with compact truth: current state, decisions, dead ends, proof, risks, and next steps.

Runtime proof

Public metrics that show the rebriefing problem getting smaller.

Measured from active TruthSpine project/chat records as of June 17, 2026.

80.13%

measured input-token savings

1,884

truth records kept

321

proofs captured

71

session syncs

93.33/100

active chat continuity

Less repeated setup at the start of every session.

Cleaner recovery when context gets compacted or interrupted.

A compact view of what is true, decided, done, risky, and next.

A practical way to keep agents aligned with the actual project state.

How it feels in use

Current truth, not a note pile.

TruthSpine is designed to preserve the small set of durable facts that make future work easier: decisions, current state, meaningful proofs, known risks, and next steps. The point is not to save every word. The point is to make the next agent action better grounded.