Agents slow down as memory notes pile up, unlike humans who speed up with practice
sujingshen · x · 2026-09-16
Citing Manling Li's framing of continual learning as "an apprentice developing expertise on the job," the author spotlights a striking finding: human testers got faster with practice, while agents generally slowed down as their memory notes grew. Merely accumulating notes can burden agents with their own memory; multi-scale abstraction—compressing experience into reusable knowledge at the right level—is what matters.
The author proposes four tests for any Personal AI claim:
- Abstraction: do failures become rules, or just another note?
- Conflict arbitration: when old and new preferences clash, who wins—and is it auditable?
- Pruning: can the agent forget/compress without losing key judgments?
- Cross-session transfer: does learned behavior survive new windows or models?
Takeaway: accumulation is not continual learning—a memory that slows you down isn't expertise yet.
More from coding & agent
- Unsloth launches desktop app to run and train LLMs locally — thisdudelikesAI · 2026-09-16
- 160 agents ran overnight: tokenmaxxing vindicated — kevinnbass · 2026-09-16
- Dev builds Chrome extension putting AI inside WhatsApp Web, seeks free beta testers — ManufacturerLast7833 · 2026-09-16
- How to Debug a RAG System: A Six-Step Framework from Retrieval to Guardrails — ashishllm · 2026-09-16
- Factory Life: Six free sessions on building with a team of AI agents, from Warp to Hermes — intellectronica · 2026-09-16
- Coding agents write code faster, but debugging them is stranger and more tedious — tengyanAI · 2026-09-16