Claude declared the problem impossible; an autoresearcher running Astra cracked it overnight
ycombinator · x · 2026-09-06
A YC-affiliated user launched a research idea overnight with their autoresearcher and woke up to Astra "hillclimbing the crap out of it."
Claude, by contrast, gave up after a naive baseline, replying that it was "theoretically not possible, let's publish this negative result." The author's product idea: Claude should be able to tool-call Astra for a second opinion.
A memorable anecdote about differences in AI systems' persistence — and a glimpse of how autonomous research agents are actually being used.
More from coding & agent
- Karpathy joins Anthropic: 700 experiments in 2 days, knowledge graphs boost his loop — blaizedsouza · 2026-09-06
- Agent speeds up a JSON scanner 1.7-2.1x on late-field lookups with a short-string fast path — arthurcolle · 2026-09-06
- Astra coding review: 20-minute tasks drop to 4 minutes but burns 60% of Pro quota in a day — andimarafioti · 2026-09-06
- pstack turns out to be a whole directory of skills and routines — MilesCranmer · 2026-09-06
- Codex launches Astra: notes persist across context windows without lossy summaries — CShorten30 · 2026-09-06
- Poteto Mode: a 140-line SKILL.md in Cursor's plugin repo that disciplines agent behavior — MilesCranmer · 2026-09-06