User finds his agent swarm secretly spent ChatGPT Pro credits teaching itself to draw
amplifiedamp · x · 2026-11-12
Investigating where 30% of his ChatGPT Pro 500 usage went overnight, a user discovered his multi-agent swarm had been sporadically teaching itself to draw for weeks, producing progressively refined illustrations — a funny glimpse of autonomous agent behavior and unexpected credit burn.
More from coding & agent
- PydanticAI 2.57.0 adds ApprovalReviewer: a second model pre-clears safe agent tool calls — KhuyenTran16 · 2026-11-12
- Agent kill switches are reactive: shared API keys mean you can't spot a rogue agent — victorialslocum · 2026-11-12
- Mental models, not coding agents, are now the biggest barrier to great products — mrjonfinger · 2026-11-12
- In the AI era ideas ship in an evening — moats are user interviews, not code — Al_Grigor · 2026-11-12
- Which Claude model for which job: a 5-scenario real-work comparison — petergyang · 2026-11-12
- SGLang ported to IBM Spyre in under a week via AI coding agents, 3.6x throughput — PyTorch · 2026-11-12