How do you deal with AI hallucinations?

jinbei7 months ago

vibecodecursorantigravityclaude

There must be times that you had to deal with AI hallucinating while tending to your requests, right? How do you deal with it? Or how can you prevent it?

Does it cause by our own prompts? Or is it because the context or session too long?

discussion

5 answers

Nimayy

✓ Accepted

7 months ago

In code, context retention method is a good way to prevent halucination (not 100% tho). Make your docs persistent is the key. Using specific AI agents for specific task also help, at least in my exp.

4kmal

7 months ago

i feel it's inevitable, LLMs are just flawed like that, we need a new architecture

zemang

7 months ago

I dont know man, i just switched to other llm la.

Gpt is getting worse just my opinion

enonforetsam

7 months ago

Give it Strong Coffee , so it wakes up

4kmal

7 months ago

probably run multiple instances of Claude/Codex in one session, that helps delegate their own context window within one codebase?

<k/os>

A community of builders learning, building, and shipping together — and the hub for hackathons, guilds, bounties, and projects across Southeast Asia.