Experiment: GPT-6 Astra wrote 75,000 lines of useless code in 35 hours
Flask creator Armin Ronacher gave GPT-6 Astra a single prompt and let it run unattended for 35 hours. The agent spent about $1,200 and 1 billion tokens, producing 79 commits and 75,000 lines of code that he called "absolutely nothing of value."
- 79 commits for $1,200 — roughly $15.50 per commit
- About 1 billion tokens and 1,400 messages between agents
- Agent code is twice as verbose and eroded as human repos on SlopCodeBench
- Strict solve rate for all tested state-of-the-art models was 0%
Read next
AI