Garry Tan finds Opus 5.5 with Openclaw oddly beats GPT-6 Astra at task completion
garrytan · x · 2026-09-27
YC President Garry Tan reports that Opus 5.5 paired with Openclaw is "strangely smarter and better at completing tasks" than GPT-6 Astra, calling the result surprising. A first-hand model comparison signal for agent workflows.
More from coding & agent
- Loop Engineering: Building a No-Prompt CI Automation That Files Jira Tickets — Pavan_Belagatti · 2026-09-28
- Anthropic engineers spend 24 minutes revealing hidden Claude Code features — jggomezt · 2026-09-28
- LocalVocal brings fully offline voice chat to local LLMs and agents on Mac — joshwhiton · 2026-09-28
- Penalizing 'wait'/'maybe' tokens lifts Qwen3.5-4B math accuracy up to +12 points — am17an · 2026-09-28
- Idea: one-click AI-ready dev environments with skills and context per project type — Shoddy_Ad1207 · 2026-09-28
- OpenScience exits beta on Product Hunt, claims #1 scientific agent with 300+ research skills — SimplyAnnisa · 2026-09-28