Agents don't need memory, they need documentation
- ID
- 31648
- Status
- summarized
- Published
- 04 Oct 2026, 1:03 AM
- Fetched
- 04 Oct 2026, 11:21 PM
- Provider
- Hacker News
- Category
- dev-community
- Original URL
- https://liao.gg/blog/agents-dont-need-memory
- Source URL
- https://hnrss.org/best
Summary
- Score
- 7.5
- Created
- 04 Oct 2026, 11:22 PM
- Tags
- Audience
- developersai_ml_learnersai_agent_usersvibe_coderssaas_founders
What happened
In a post dated October 3, 2026, Kevin Liao argues that every "memory" plugin on the market shares one architecture: chunk session transcripts into snippets, embed them in a vector store, inject the top 5 on every prompt, and optionally give the agent a search tool. He lists five failure modes — similarity is not correctness, snippets lose context, the past is treated as truth even as the codebase changes, agents can't search for what they don't know, and a store of e.g. 10,000 embeddings in SQLite is unauditable. His claim is that agents need versioned, auditable documentation rather than recall, and the Hacker News thread drew 261 points and 144 comments.
Why it matters
If you're paying for or building a memory plugin, this says swapping vendors won't fix the failure mode, because the transcript-to-vector-to-top-5 pipeline is the same everywhere — the difference is only extra token-burning layers like rerankers, dedup daemons, or overnight 'dreamer' rewrites. The practical decision it pushes: before adding another memory tool, check whether the agent can instead read a versioned doc that states what is current, and whether you can audit which snippets are stale or never retrieved. Note this is an argument from experience, not a benchmark — no accuracy numbers are given, so treat the 'none of them reliably work' claim as a hypothesis to test on your own repo.
Discussion angle
Take one agent you've wired a memory plugin into and count what's actually in its store: how many snippets, how many ever get retrieved, and how many describe code you've since changed. Then ask the room whether replacing that store with a maintained doc file would have changed any output — and whether the 144-comment thread's counterarguments (that docs also rot, and that agents won't read them unprompted) hold up.