# Are your agents actually learning, or are you just getting better at maintaining them? — retained record

> Retained reddit record from reddit:AI_Agents.

Canonical: https://highsignal.app/data/records/265d2ef67e2e2d8b

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:AI_Agents)  published 2026-10-24 09:11 ingested 2026-10-25 02:35

[original source ↗](https://www.reddit.com/r/AI_Agents/comments/1xe3yaq/are_your_agents_actually_learning_or_are_you_just/)

## retained text



```
I want an agent that gets better at working with me without me constantly updating its prompts, skills and memory.



Give it an unfamiliar customer problem. Let it talk to users, build a small service, get it into a pilot and improve it over months. I shouldn’t have to design a separate workflow for every step.



The interesting part is whether what it learns carries over. If it learns to question misleading evidence while debugging, does that help it question a claim in a customer interview? Can it also recognize when that comparison doesn’t apply?



I experimented with persistent memory in my Cortex project during Anthropic’s Claude Code hackathon. Some comparisons improved. Others favored the baseline. So I’m not convinced that just adding more memory solves this.



I think adapting model weights could help with skills the agent uses repeatedly. But I’d want to compare that against a strong memory and skills setup, rather than assume it’s better.



The test I care about is pretty simple: give an experienced agent and a fresh copy unfamiliar work. Does the experienced one get better results, need fewer repeated corrections and require less maintenance from me? Does it retain what it already learned?



Has anyone tested this across genuinely different tasks? I’m curious what actually transferred and what didn’t.
```



document fetched 2026-10-25 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1xe3yaq/are_your_agents_actually_learning_or_are_you_just
