Threadbaire

Threadbaire

Open method for keeping memory and context together when working with AI. File conventions that any agent can read. Model-agnostic, portable, user-controlled.

What it is

Markdown file conventions for AI project memory. Two files — a Core brief and a running Addendum with receipts (who · why · source · date · model). Your context travels between tools and models because it lives in files you control.

Includes RundownAPI — an open spec for API endpoints that teach AI how to integrate. One URL, no SDK, no setup. The endpoint describes itself, and the agent builds its own connector. Tested with Claude Code, which read a single RundownAPI endpoint and wrote its own integration without pre-built tooling.

The argument

The thesis — The Software Collapse Has Already Happened — argues that AI has collapsed traditional software value capture while platforms race to capture the decision layer through proprietary context graphs. Threadbaire is the counter-position: your reasoning stays in files you own.

The blog extends the thesis with structural analysis of AI industry dynamics, agent infrastructure, and what comes after per-seat software.

Repositories

• threadbaire/method — Markdown templates (CC BY-SA 4.0)
• threadbaire/rundownapi — Open spec for AI self-integration endpoints (CC BY-SA 4.0)
• threadbaire/server — Reference implementation, frozen at v1 (MIT)

Author

Lida Liberopoulou · Independent research · Available for advisory, research collaboration, and expert consultation