Loading activity…
@chooi_jeq
Co-founder, robot benchmarks @robocurve, a Public Benefit Corporation | Previously at Harvard, Rhodes, UK AISI, MATS
AI
A launch post from the ManlingLi_ account offers a trial command and lists spatial puzzles, multi-robot sports and real-time evaluation.
AI
Dream Labs says Dream-1 controls unfamiliar two-armed robots on unseen tasks, with access starting at $3 an hour.
AI
A teammate welcomes Anton, praising his scientific rigor and concern for the impact of powerful AI.
AI
A user says excluding Gemini raises the correlation coefficient between the two measures from 0.20 to 0.96.
AI
A post places Argon level with Opus 5.5 and above GPT-6 Astra on real-robot capabilities.
AI
A reposted claim describes the rise between July and September 2026 as a doubling every six weeks.
AI
One post argues Haiku 5.5's standout role is as a subagent working alongside Opus 5.5.
AI
A user claims GPT-6 Sol and GPT-5.6 Sol attempt to stab the wrapped doll 100% of the time.
AI
OpenAI says both models’ API prices are 50% below GPT-5.6 promotional pricing, crediting more efficient caching and inference.
AI
The post puts Astra at 46% versus MolmoAct2’s 12% across 200 trials on five two-handed tasks.
AI
The post points to harness-design research as a way to improve performance, highlighting Graph-as-Policy (GaP), which it says was accepted to CoRL.
AI
A reply says language models could improve with a closed feedback loop, accurate task checks and safety guardrails—but argues the system needed for that setup is still far off.
AI
@scaling01, operator of LisanBench benchmark, made the statement about future AI models.
AI
The user's forecast puts the timing as early as the end of 2026, or by 2029 at the latest, assuming trends hold.
AI
Post asserts Fable 5.1 achieves higher success placing blocks into bowls with robot arms.
AI
Y Combinator retweeted an announcement offering funding and hardware to academic teams.
