AI ResearchMusic & AudioAgents5 min reading time

OpenAI tops ARC-AGI-3 📈, GPT-5.6 efficiency ⚙️, AlphaFold team dissolution 🧬

TLDR AI
Read full post
OpenAI's GPT-5.6 Sol model initially scored low on the ARC-AGI-3 benchmark, but enabling retained reasoning and compaction settings tripled its performance and reduced output tokens sixfold. Meanwhile, SpaceXAI released Grok Voice Think Fast 2.0, and Google's Lyria 3.5 music model launched with improved musicality and vocals. Additionally, researchers introduced efficient CPU-friendly long-context encoders and explored visual prompt engineering for video models.

More in AI Research

Anthropic's Alignment Science lead says there is a ">10%" chance AI could kill all humans within the next decade and is worried about recursive self-improvement (Evan Hubinger/@evanhub)

Covered by 9 sources
AI Research4 min read

Suno trained its v6 AI music models with help from Warner and BMG

Covered by 5 sources

Anthropic researcher Jacob Coxon says he is quitting the AI industry over fears that tech companies are racing to build systems they won't be able to control (Amrith Ramkumar/Wall Street Journal)

Covered by 11 sources