10 articles on ai research.
AI mathematics discovery is moving from benchmarks to research. Here is what the Navier-Stokes claim proves, what remains disputed, and what builders should do.
OpenAI’s Navier-Stokes AI proof claim could reshape scientific agents. Here’s what is verified, what Bell means, and what creators should do.
Gemini Notebook adds agentic research, code-powered analysis and downloadable outputs. Learn how creators and teams can use it responsibly.
DiffusionBlocks trains neural networks one block at a time, aiming to cut GPU memory needs without sacrificing model quality or flexibility.
Claude’s Riemann Hypothesis advance did not solve the famous conjecture—but its 67.2% bound offers a new model for AI-led research.
Hybrid imitation learning combines motion tracking and adversarial training to create lifelike virtual athletes that adapt to unseen parkour courses.
The Free Transformer adds learned latent variables to decoder LLMs. Here’s how it works, what benchmarks show, and why builders should care.
Neural computers turn AI into a learned runtime for software. See how early prototypes work and why reliable state remains the central barrier.
Evolution strategies for LLM fine-tuning are back. Learn how ES-at-Scale and EGGROLL challenge RL for outcome-based model training.
LLM character counting reveals how Claude 3.5 Haiku tracks line length through curved internal representations—and why it matters for AI safety.