US startup advertises ‘AI bully’ role to test patience of leading chatbots
TL;DR
California startup Memvid is offering $800 per day for an 'AI bully' role – someone who spends eight hours probing chatbots for inconsistencies, hallucinations, and memory failures.
Key Points
- The job involves deliberately challenging AI systems, surfacing contradictions, and documenting the limits of current large language models.
- The listing highlights a genuine problem: chatbots forget context, fabricate facts, and often do so with misplaced confidence.
- Rate: $100 per hour – for professionally structured antagonism.
Nauti's Take
$800 a day to pick fights with a chatbot sounds absurd – but it is symptomatic of where the AI industry actually stands. If language models were as reliable as their makers claim, nobody would need paid bullies.
Memvid at least makes transparent what many companies do internally: human stress-testing, because automated evals alone do not cut it. The real question is whether eight hours of manual provocation changes anything fundamentally, or whether this is mostly marketing for a startup that itself builds memory tools for AI.
Sources