Another research from @_doubleAI_ Most RL objectives for language models maximize average reward per generation. They do not explicitly reward maintaining several distinct, high-quality solutions, so output diversity can shrink during training—limiting the gains from best-of-k sampling. ArgMaxRL directly optimizes the expected quality of the best among k samples for continuous rewards. This better aligns training with test-time scaling: generate multiple candidates, verify them, and select the best. The result could be models that preserve useful diversity and benefit more from additional inference compute. doubleai.com/research/argmax…

WarpSpeed is @_doubleAI_ 's self-improving, formally verified AI system. We've used it for kernel optimization (cuGraph), performance engineering (SOL-execbench), optimizing Python libraries like Pandas, and quantum circuits for ECDSA. Now: hardware design. NVIDIA's CVDP covers the digital-design stage — write RTL from a spec, complete it, modify it, debug it, and optimize it. The optimization split is the one we focused on: given existing RTL and a spec, produce a functionally identical design using fewer cells. Cell count drives area and complexity, so it's scored as % reduction against the original. WarpSpeed hits 82.5% there, and solves 98.5% on the design split. Guaranteeing correctness is the harder problem. A design that drops its snapshot registers and reads live wires passes every CVDP test — the tests hold inputs fixed — while behaving differently from the spec. Claude Code w/ Fable scores 80.1% on average over all problems in the benchmark, but its solution on 16.2% of the problems is invalid. WarpSpeed solutions were correct on 100% of the problems, because every candidate has to discharge a formal contract derived from the spec and original RTL before it's accepted. For more details see doubleai.s.gy/doubleAI-RTL

1/ Since WarpSpeed topped NVIDIA's SOL-ExecBench (99% of its hardest kernels, 3x faster than NVIDIA's own baselines), the question I keep getting is the right one: that is a benchmark. Can it do the same on the workloads I actually run? We are about to find out, in the open.

DoubleAI’s AI system just beat a decade of expert GPU engineering WarpSpeed just beat a decade of expert-engineered GPU kernels — every single one of them. cuGraph is one of the most widely used GPU-accelerated libraries in the world. It spans dozens of graph algorithms, each written and continuously refined by some of the world’s top performance engineers. @_doubleAI_'s WarpSpeed autonomously rewrote and re-optimized these kernels across three GPU architectures (A100, L4, A10G). Today, we released the hyper-optimized version on GitHub — install it with no change to your code. The numbers: - 3.6x average speedup over human experts - 100% of kernels benefit from speedup - 55% see more than 2x improvement. But hasn’t AI already achieved expert-level status — winning gold medals at IMO, outperforming top programmers on CodeForces? Not quite. Those wins share three hidden crutches: abundant training data, trivial validation, and short reasoning chains. Where all three hold, today’s AI shines. Remove any one of them and it falls apart (as Shai Shalev Shwartz wrote in his post). GPU performance engineering breaks all three. Data is scarce. Correctness is hard to validate. And performance comes from a long chain of interacting choices — memory layout, warp behavior, caching, scheduling, graph structure. Even state-of-the-art agents like Claude Code, Codex, and Gemini CLI fail dramatically here, often producing incorrect implementations even when handed cuGraph’s own test suite. Scaling alone can’t break this barrier. It took new algorithmic ideas — our Diligent framework for learning from extremely small datasets, our PAC-reasoning methodology for verification when ground truth isn’t available, and novel agentic search structures for navigating deep decision chains. This is the beginning of Artificial Expert Intelligence (AEI) — not AGI, but something the world needs more: systems that reliably surpass human experts in the domains where expertise is rarest, slowest, and most valuable. If AI can surpass the world’s best GPU engineers, which domain falls next? For the full blog: doubleai.com/research/double… CuGraph: docs.rapids.ai/api/cugraph/s… Winning Gold at IMO 2025: arxiv.org/abs/2507.15855 Codeforces benchmarks: rdworldonline.com/openai-rel… @shai_s_shwartz post: xitter.cf/shai_s_shwartz/status/… From Reasoning to Super-Intelligence: A Search-Theoretic Perspective arxiv.org/abs/2507.15865 Artificial Expert Intelligence through PAC-reasoning arxiv.org/abs/2412.02441