Jim Fan's lab open-sources ASPIRE robotics framework, benchmarks hit ~99% overnight with Astra
DrJimFan · x · 2026-10-17
NVIDIA's Jim Fan announces the open-source V1.0 of ASPIRE, an agentic robotics framework with full real-robot skill discovery and self-improvement.
- First introduced in June, before frontier models could reliably do robotics
- The lab says switching the backend to Astra saturated all benchmarks to 99% overnight
- Opus 5.5 slightly edges out Astra on their robot simulation bench; Sonnet 5.5 hits the Pareto frontier of cost vs. capability at $12/task
Links to code and site in the thread; the team calls it a demonstration of the "Physical Bitter Lesson."
Related event: NVIDIA Open-Sources ASPIRE, a Self-Improving Robot Framework(3 posts)→
More from coding & agent
- shadcn reflects: it's only been a year since Claude Skills launched — shadcn · 2026-10-17
- ClinOpsBench: 15 AI agents tested on 4-month simulated clinical trial oversight — ycombinator · 2026-10-17
- PR Lens matches agent PR screenshots to architecture diagrams in ~2s, 2k GitHub stars — nevrekaraishwa2 · 2026-10-17
- Cohere research: agent teams hit frontier performance at 30% of the cost — nouhadziri · 2026-10-17
- Four agent frameworks shipped the same fix in one week: context compaction vs approvals — HelpfulCollar1070 · 2026-10-17
- Pipecat's open PhoneLLM Max Alpha 2 tops voice-agent bench at 83%, beats Sonnet 5.5 — solyarisoftware · 2026-10-17