Notes on agent memory, continuity, and building with Stele
How-tos, comparisons, benchmark findings, and engineering notes from the team. For people deciding how their agents keep and share project context.
Why markdown rot breaks agent memory
Three failure modes of repo-local agent notes, and what Stele’s shipped lifecycle does about each one. No unique-staleness claim.
Read post
When a fact stops being true
How Stele retires a reversed decision without deleting the story: supersession, reasons, and why the replacement always arrives with it.
Read post
A file, a wiki, or a shared ledger?
Answering the most common objection: when Notion, Obsidian, or a markdown file is enough, and when you need a maintained project record.
Read post
Switch agents without losing the project
Plan in one tool, build in another, hand off on a quota fill: how a shared project record keeps continuity across harnesses.
Read post
How an agent session uses Stele
Context in, work claimed, knowledge written back: the day-to-day loop around the shared record, as it ships today.
Read post
Stele vs the usual ways to give agents memory
CLAUDE.md, Notion, IDE memory, Mem0/Zep, Letta Code: what each is for, and when a shared project ledger is the better fit.
Read post