An agent can give a good first answer and still fail when the user corrects it.
This is a dev post classified by AGK Intelligence as Testing & observability (a tool drop), kept by the Dev Radar because it carries real work, not commentary.
An agent can give a good first answer and still fail when the user corrects it. Scenario, from LangWatch, tests agents through conversations with simulated users. You describe the situation, and a user simulator generates messages and follow-ups. You can also write specific parts of the conversation yourself, then let the simulation continue. One example tests an agent that mistakes a river bank for a banking question. After the user corrects it, the test checks whether it follows the clarification or keeps giving banking advice. You can check behaviour along the way, such as whether the a
Posted by Sandhya (1.4k followers) 1 h ago · 0 likes · 23 views · view the original post on X. Kept by the Dev Radar as Testing & observability.
More dev work like this
- Most of our tests reuse a single mysql container, but some tests need isolated databases… — @chronark
- Your CI jobs deserve a clean start. — @novita_labs
- CI pro tip: — @DanielLockyer
- build chat agents you can trust in production — @langfuse
- this one is big 👀 — @blazejkustra_
- TheGrill made it into TensorFold’s benchmark docs. — @plotarmordev
- A typical GitHub repo uses `ubuntu-latest` for their CI runners, but I wonder how many… — @DanielLockyer
- If you're working on making your CI faster and want your agent to visualize the timeline… — @DanielLockyer
Every post is read and classified by AGK Intelligence: what it is, which market it belongs to, and whether the link is a real tool. 31.3k posts from 5k X accounts over the last 21 days, 9.1k tools, 12 markets. New posts are collected several times an hour and the full ranking is recomputed every hour — last update 2026-11-02 20:07 UTC. Full method.