Best "Brain" for Agents Is Just Versioned Folders of Markdown Files
murdockq
35 points
8 comments
May 14, 2026
Related Discussions
Found 5 related stories in 87.4ms across 8,303 title embeddings via pgvector HNSW
- Show HN: Agent Kernel – Three Markdown files that make any AI agent stateful obilgic · 40 pts · March 23, 2026 · 64% similar
- Show HN: Brainio – Markdown notepad that turns notes into visual mind maps havlenao · 20 pts · May 05, 2026 · 61% similar
- New Research Reassesses the Value of Agents.md Files for AI Coding noemit · 19 pts · March 08, 2026 · 59% similar
- Your Agents.md is probably too long jlong · 14 pts · March 18, 2026 · 58% similar
- Parallel coding agents with tmux and Markdown specs schipperai · 139 pts · March 02, 2026 · 56% similar
Discussion Highlights (7 comments)
jfreds
I’m confused: is it just markdown files in git? Or does the hybrid graph+semantic layer matter? If the latter is true, the title is just clickbait right
skiing_crawling
How does git replace a vector db search exactly? They are orthogonal. Are you gonna burn a million tokens every time you wanna find some relevant files?
GrinningFool
This isn't even well-slopped slop.
TranquilMarmot
This should be common sense and immediately obvious to anybody who has spent more than a few hours with a coding agent.
Kim_Bruning
I think just one useful thing is mentioned: put your md files in git and then have a hybrid fts/vector search across them actually does work better? Not a very surprising conclusion. Doesn't need that much text to explain, does it?
knighthack
I really don't get having a ton of MD files lying around. And the possibility of having to edit a thousand MD files when the metadata frontmatter changes. A single SQLite database implements columns/metadata handling, and comes baked-in with FTS and BM25 ranking too.
zwaps
This reads like an AI writing to a point an author wants to make without there being evidence. It is completely incoherent. Apparently we just need markdown and git, but also a knowledge graph and pgvector which accounts for most of the performance. We don’t need semantic search, because we use… hybrid search (semantic search plus bm25)??? Really bad look for an AI consulting company this.