Agent Memory Is Still Just a Markdown Folder
steipete · x · 2026-07-20
The core point reshared by the author is that his agent memory is still just a folder of Markdown files, but as models iterate, its performance keeps improving with almost no extra modifications needed.
The reason is that models have undergone massive post-training for coding scenarios, making them increasingly adept at navigating file trees: using grep, reading files, and tracing context along links. In contrast, vector databases or memory APIs cannot inherit these capabilities. At around 900 files currently, this simple solution is sufficient, and he can directly open the vault to see everything the agent remembers.
More from coding & agent
- Hermes Agent rewrite proposal applies RIA and Logic Bus rules — Promptmethus · 2026-07-22
- Devin adds e2b sandboxes for remote agent execution — badphilosopher · 2026-07-22
- LangChain adds LangSmith tracing for voice agents across four frameworks — LangChain · 2026-07-22
- Hermes Agent Refactoring Proposal: Decoupling via Event Bus and Monorepo Slicing — Promptmethus · 2026-07-22
- ty now reads Pydantic config keywords and field metadata — charliermarsh · 2026-07-22
- Pensar Launches AI Security Agent to Autonomously Discover and Patch 0-Days — andriy_mulyar · 2026-07-22