LangChain's Chase on why agent memory never sticks: deciding what to remember is app-specific
blaizedsouza · x · 2026-09-15
LangChain founder Harrison Chase responds to Cursor CEO's claim that memory is the most under-invested space, explaining why it's been a hot topic for two years without a breakout product.
- Updating and using memory must be tightly integrated with the harness, making a standalone memory product hard
- The hard part isn't storage or querying but the logic of deciding what to remember — often (a) application-specific and (b) just prompting, so hard to productize
- Memory matters for repeated tasks, not general-purpose agents; high-usage agents aren't doing the same thing over and over
- Memory hasn't proven useful for coding agents, where most successful agent patterns emerge first
More from coding & agent
- ASTRA + Thrixel Build an Entire 3D Rail Shooter From a Single Prompt — RanaHanocka · 2026-09-15
- One prompt builds a full 3D sci-fi rail shooter, playable in browser — RanaHanocka · 2026-09-15
- Same feature, 3 models: $4 Claude vs $0.50 Tencent Hy3 — the cheapest won — tejasghutukade · 2026-09-15
- Stripe asks AI agents operators for feedback and feature requests — jeff_weinstein · 2026-09-15
- AI agent finds $3,500/year car insurance savings in 5 minutes, buys it and cancels old policy — garrytan · 2026-09-15
- Pokee Isaac MCP now runs inside OpenCode, generating complete artifacts in one pass — Kyrannio · 2026-09-15