The Shift to Test-Time Compute
For a decade, progress meant scaling training. Reasoning models flipped the axis: spend compute at inference, let the model deliberate, and accuracy climbs on the hardest problems.
Signal, not noise
A clear read on the research shaping AI.
Careful, cited analysis of the latest papers in machine learning — written for the people who build with them.
Featured
What Makes an AI Agent Actually Work
An agent is a loop: the model observes, decides, acts through tools, and reads the result. The hard parts are error recovery, context, and knowing when to stop — not raw intelligence.
Latest
All topics →
Mixture-of-Experts, Explained
How a trillion-parameter model can run at the cost of a much smaller one.
World Models and the Race for Video
Generating video is really about learning how the world behaves.
Why Small Models Are Suddenly Good Enough
The frontier gets the headlines. Small models are quietly winning production.
RAG Is Not Dead — It Grew Up
Long context didn't kill retrieval. It changed what retrieval is for.