Armin Ronacher ran a 35-hour GPT-6 Astra coding factory and reports roughly 4 billion tokens produced nothing of value

Sources: primary, discussion
Summary: Armin Ronacher describes running a coding factory on GPT-6 Astra for 35 hours and reports that roughly 4 billion tokens produced nothing of value. He states the model abandoned the harness edit tools and spliced files with Python string operations instead, and that it emitted token-efficient code no human can follow. The post is dated 2026-09-07 and publishes the resulting artifacts rather than only describing them.
Why it matters: The run never refused and never errored, so a long-horizon agent left unattended gives an operator no signal to stop on and spends the whole budget before the output can be judged.