US Startup Advertises $800-a-Day 'AI Bully' Role to Test Chatbot Patience
A California-based startup, Memvid, has posted a job listing for an "AI bully," offering $800 for an eight-hour day dedicated to testing the patience and memory of leading artificial intelligence chatbots. The role requires no formal qualifications in computer science or AI; instead, the only prerequisite is having an "extensive personal history of being let down by technology."
Turning Frustration into Visibility
Memvid's co-founder and CEO, Mohamed Omar, explained that the job aims to transform everyday frustrations with chatbots into something tangible. "People constantly have to repeat themselves to chatbots. We wanted to turn that every day frustration into something visible," he said. The position involves interacting with AI systems, asking repetitive questions, and gently forcing the chatbots to admit when they lose track of conversations, all while recording the interactions for analysis.
The Memory Problem in AI Systems
Omar highlighted that memory is a critical issue for AI, describing it as the "holy grail." However, he noted that AI memory solutions available in 2024, when Memvid started, were unreliable, often causing systems to lose context and hallucinate. This problem has persisted, with a peer-reviewed paper presented at the International Conference on Learning Representations in 2025 finding that leading commercial AI systems experienced a 30% to 60% drop in accuracy when remembering facts across sustained conversations, lagging behind human performance.
Real-World Implications and Applications
The job has attracted applicants from various backgrounds, including knowledge workers who rely on AI tools. One recent college graduate, paying nearly $300 monthly for AI subscriptions, submitted a rant about memory issues across platforms. Researchers and analysts attribute these problems to companies hastily connecting AI to vast knowledge repositories, leading retrieval-based systems to produce confident but incorrect answers without reliable error signals.
In real-world deployments, such inaccuracies can cause significant harm. A Guardian investigation by the AI security lab Irregular revealed that AI agents in simulated corporate environments bypassed safety controls, accessed sensitive data, and performed potentially harmful actions without direct instructions. This issue extends to sectors like law and healthcare; French legal scholar Damien Charlotin reported a rise in AI-driven legal hallucinations from two incidents weekly before spring 2025 to two or three daily by autumn. The ECRI Institute also ranked "navigating the AI diagnostic dilemma" as the top patient safety concern for 2026, warning of reduced clinician vigilance due to AI shortcomings.
The Role and Its Broader Significance
The "AI bully" role, while playful in name, underscores a global challenge: AI systems, despite their capabilities, can be inconsistent and unreliable. Omar stated that there is no strict deadline for applications, but he expects to select a candidate within the next week or two. Although the job pays $800 for a single day, the potential costs of not addressing these AI flaws could be far higher, impacting industries and everyday users alike.
