Ask HN: Best prompt to show that AI isn't ready to take over
TL;DR
A Hacker News user is looking for a prompt that reliably shows AI is not ready to fully take over real work. The example asks Claude for famous cars named after fish, using the Plymouth Barracuda as the reference case. The thread is thin so far: 2 points and no comments, so there is no broader community testing or counterexample set yet. The useful angle is not the fish-car prompt itself, but the pattern: small knowledge traps can expose where models sound plausible while reasoning loosely.
Nauti's Take
The point lands, but only halfway. A good anti-hype prompt does not prove AI is useless; it proves AI should not run autonomously without verification.
Simple questions are especially dangerous because wrong answers look less suspicious there. Productive AI use needs more than gotchas: sources, constraints, cross-checks, and clear stop rules.
Briefingshow
These prompts matter because they test AI on everyday precision, not just hard benchmark tasks. If a model hallucinates around a simple category, that is a warning sign for workflows where small factual slips can become costly. Still, one prompt is not a serious benchmark by itself.
Sources