What if code could think while it runs? Meet Maxwell: the world’s first Code Execution Model. A new species of AI that lives inside the machine. We believe it’s the key to unlocking 100% of compute. Requests for early access are open, link below.

3/ Harder question: what if both players keep answering check with check? The longest known sequence is 54 consecutive checks, in a composed position by Alexey Khanyan (2013).

2/ The answer: 63. Every one of them gives check to Black’s king. That beats Norbert Ringeltaube’s record of 60, and Double Agent proved that 63 is the maximum possible. The position can be reached in a legal game.

1/ What is the maximum number of legal moves in a chess position where every one of them gives check? The record on that question had stood since 1968. Double Agent, our AI research system, just broke it. How many checks can you count here?

2/ A graph's girth is the length of its shortest cycle. Large girth means a graph looks like a tree for great distance around every vertex. Such graphs are key to constructing error-correcting codes. Since 1988, the best explicit graphs, due to Margulis and to Lubotzky, Phillips and Sarnak, had girth (4/3) log n, against a theoretical ceiling of 2 log n.

1/ Double Agent, our agentic system, has achieved a breakthrough result on one of the most central problems in extremal graph theory. It constructed explicit regular graphs with girth (8/5) log n, the first improvement in 38 years, and a leap towards the theoretical maximum.

Can AI reach superoptimisation across the internet, at scale? What if every single workload you ran was tailored to your exact hardware configuration? Gal Beniamini (@laginimaineb) is speaking about it at @NVIDIAGTC Berlin, Thu 22 Oct, 15:00 CEST.

To power our research, we need compute. At doubleAI, we decided compute ownership is a long-term goal. We began modestly a year ago with a cluster of 96 B200 GPUs. Now we're building a cluster of 1,024 B300 GPUs, arranged in 128 DGX servers, totaling 2 MW of power.

3/4 Why that matters: most real rewards aren't yes/no. Ten CUDA kernels can all be correct but run 1.2x to 8.5x faster. Pass@k only sees "correct," not which one is fastest. ArgMaxRL optimizes directly for the best sample.

1/4 On a hard problem, an AI's first answer is rarely its best. The gains come from generating many tries and keeping the best. But standard RL collapses a model onto one answer, so more samples don't help. ArgMaxRL fixes that for continuous rewards, not just pass-or-fail.