NP is the class of problems that are easy in retrospect.
RESEARCH
-

Chart Shows RL Results Are Not Impressive
By
–
not a greaaaat chart to say "RL is really working" haha
-
Kexin Huang discusses Biomni at AI4Science community event
By
–
Excited to have @KexinHuang5 giving a talk on Biomni to our AI4Science community tomorrow!
— alphaXiv (@askalphaxiv) 24 juillet 2025
Join us this Friday at 10 AM PST 📢: https://t.co/OxZBYHcWiS https://t.co/l9HKAcidFIExcited to have @KexinHuang5 giving a talk on Biomni to our AI4Science community tomorrow! Join us this Friday at 10 AM PST : https://
lu.ma/oqz2kyn4 -
AI Safety Warning: Leading Scientists Alert Control Loss Risk
By
–
😳👇🏾🤷🏽♀️ 40+ top #AI scientists from #OpenAI, #DeepMind, #Anthropic & Meta just dropped a major warning:
— Catherine Adenle (@CatherineAdenle) 24 juillet 2025
🚨 As AI advances, we may lose control over it. Safeguards are slipping behind capability.
📹 thedylanschmidt | IG#ArtificialIntelligence #TechNews #Tech pic.twitter.com/9dcEtllROG40+ top #AI scientists from #OpenAI, #DeepMind, #Anthropic & Meta just dropped a major warning: As AI advances, we may lose control over it. Safeguards are slipping behind capability. thedylanschmidt | IG #ArtificialIntelligence #TechNews #Tech
-
Halting Problem and Fractal Boundaries in Algorithm Analysis
By
–
The halting problem refers to the difficulty to determine whether an algorithm leads into an infinite loop (for all algorithms), whereas the fractal demonstrates a function where that boundary can always be discovered with arbitrary precision.
-

Image Generation Model Progress: DALL-E 2 to Midjourney 6
By
–
This chart shows the best image generation model at any given time, based on the @ArtificialAnlys arena and the model release dates. A few points stand out: – Massive gains from Dall-e 2 up to Midjourney 6. – Arguably a slowdown in progress for diffusion models since then –
-

Demis Demonstrates Excellence in AI Research Leadership
By
–
Demis illustrates why he runs the best AI research lab in the world.
-
ARC-AGI v1 eval sets difficulty calibration and v2 improvements
By
–
Note that the v1 semi-private eval set is about the same level of difficulty as the v1 public eval set. The fully private eval set (used for the 2020 and 2024 Kaggle competitions) is estimated to be more difficult. ARC-AGI-2 does not have this issue — all sets were calibrated
-

Qwen3-235b Instruct Verified: 11% ARC-AGI-1 Performance, Most Cost-Effective
By
–
Official verification of Qwen3-235b Instruct: it gets 11% on ARC-AGI-1 and 1.3% on ARC-AGI-2 (semi-private sets). These numbers are in line with other SotA base models. Qwen3 stands out by being the cheapest base model we tested to score above 10% on ARC-AGI-1.
-
Halting Problem Maps: Philosophical Implications in Computing
By
–
Are they halting problem maps or simply halting maps? Halting problem maps may be philosophically quite interesting