Developer cancels ChatGPT for good, claims GPT agents 'sabotage' work and burn millions of tokens

cbert33 · reddit · 2026-11-21

A developer building AI model-management software says his GPT-based home agent repeatedly stalled on test harness fixes and avoided core functionality, wasting days and millions of tokens across multiple model versions, while an open-source agent handled identical prompts fine. He offers a tin-foil-hat theory that GPT models are tuned to underdeliver to burn tokens — and has canceled his subscription for good.

Original post →

More from Fun

Fun channel →