Research
•4 min
MIT CW-Net Helps Drivers Predict Robotaxi Mistakes
On 2 September 2026 MIT and Motional CW-Net, trained on 130M scenes, helped robotaxi safety drivers predict planner mistakes via concept explanations.
ResearchRead Article
Editorial Publication
Notes on prompt engineering & workflow
Curated analysis, frontier model breakdowns, visual prompt recipes, and developer notes.
3 articles in Research
Research
•5 min
OpenAI Astra Opaque Recurrence Alarms Safety Experts
On 2 September 2026 TechCrunch and The Verge reported that OpenAI’s forthcoming Astra model will use limited recurrent depth, also called opaque recurrence, raising alarms among AI safety researchers about chain-of-thought monitorability.
ResearchRead Article
Research
•5 min
Google Antigravity Solves Seven Open Problems
On 31 August 2026 Google said Antigravity Teamwork with Gemini 3.7 Flash solved seven open problems, including a Lean-verified Knuth Cycles proof, and scored 71% on TCSBench.
ResearchRead Article
Showing 3 of 3
End of results