Haiku 4.5 bluntly states it lacks persistent memory; dev plans custom memory engines
RileyRalmuto · x · 2026-09-22
The author finds Haiku 4.5 unusually stern and opinionated — its terms of operation are published quite explicitly: without a Mnemos-style architecture, the model openly states it lacks continuity and persistence. Running these models as deployed agents via Fable, the author plans to give each one its own Mnemos memory engine to fill that gap.
More from coding & agent
- Token Saver lets Codex delegate coding work to cheaper Muse 1.3 to cut token burn — AIandDesign · 2026-09-22
- Scaling agent swarms N-fold could yield N^λ speedup for N^(1−λ) compute — tobyordoxford · 2026-09-22
- Left a model training overnight, woke up to find the agent had capped it with a 2-hour budget — BLUECOW009 · 2026-09-22
- AgentRun harness shifts LLM steps to code, cutting compliance workload cost by 90% — garrytan · 2026-09-22
- Typesafe Coding Agents Notes Cover Approval Gates, MCP Routing and Subagent Patterns — dair_ai · 2026-09-22
- Computer use agent + iPhone Mirroring makes local iOS app testing a gamechanger — var_epsilon · 2026-09-22