first-hand human knowledge

Attention is all that matters.

We read the podcasts, essays and interviews — and hand you the arguments: who claims what, who rebuts, and the original voice one click away.

Banning general-purpose computation in favor of specialized anal AI agents' success rate on end-to-end, economically valuable fre Slowing or banning the open-source AI model ecosystem is an effe Frontier AI agents' reliability on long-horizon, multi-step, mul

Popular moments

This week This month All time

all topics benchmarks · 17 claude-code · 9 automation · 7 sqlite-utils · 7 policy · 7 agents · 7

Fact◆Article

Fable's AI system autonomously wrote a CUDA megakernel that achieves an 18.71x speedup over an optimized PyTorch baseline using a single cooperative kernel launch per token, beating every other frontier model's multi-kernel approach on KernelBench-Mega.

Fable wrote the first megakernel submitted to KernelBench-Mega, achieving an 18.71x speedup with a single kernel launch per token, versus 4-14 launches for competing models, signaling rapid AI progress on tasks central to automating AI R&D itself.

Jack Clark · Import AISupports · 1 Extends · 1

Data◆Article

Fable's AI system wrote the first genuine and fastest megakernel ever submitted to KernelBench-Mega, hitting an 18.71X speedup with just one cooperative kernel launch per decoded token, versus rivals that needed 4 to 14 separate launches.

Fable's AI-written CUDA megakernel achieved an 18.71x speedup on KernelBench-Mega using only one kernel launch per token, beating attempts by Claude Opus 4.8, GLM-5.2, and GPT-5.5, signaling progress toward AI automating its own R&D.

Jack Clark (Import AI) · Import AI

Data◆Article

Fable's AI-written megakernel is the fastest and most efficient solution ever submitted to KernelBench-Mega, using a single cooperative kernel launch per token where every other high-scoring entry needed 4 to 14 — a sign AI is closing in on the R&D tasks that underlie recursive self-improvement.

Fable's Cuda-written megakernel hit an 18.71X speedup over an optimized PyTorch baseline, beating Claude Opus 4.8, GLM-5.2, and GPT-5.5 on the same benchmark, and did so with a single kernel launch per token versus competitors' 4-14.

Jack Clark (Import AI) · Import AIExtends · 1

Claim◆Article

Hy3 outperforms similar-size models and rivals flagship open-source models that have 2-5x as many parameters.

Tencent claims Hy3 beats models of comparable size and is competitive with much larger flagship open-source models.

Tencent Hy Team · Simon Willison's WeblogMechanism · 1

Data◆Article

Sonnet 5 requires roughly 3x the agentic turns of Sonnet 4.6 on benchmarks like AA-Briefcase and GDPval-AA, and at max effort it takes around 6x more turns than at low effort.

Artificial Analysis found Sonnet 5 consumes far more agentic turns than its predecessor—about 3x more than Sonnet 4.6 on AA-Briefcase/GDPval-AA, and up to 6x more at maximum effort versus low effort—underlying the efficiency concerns that dampened the launch's reception.

Artificial Analysis · Latent Space

DataArticle

AI models' ability to complete real, economically valuable freelance projects end-to-end more than quadrupled in under eight months, rising from 2.5% success in October 2025 to 16.1% in July 2026.

The Remote Labor Index shows frontier AI success on real freelance projects jumping from 2.5% to 16.1% in under a year, with Fable 5 leading GPT-5.5 and Opus 4.8.

CAIS and Scale researchers · Import AIExtends · 1 Supports · 1

DataArticle

AI systems' ability to complete end-to-end online freelance work success has more than quadrupled in under eight months, from 2.5% at the Remote Labor Index's October 2025 launch to 16.1% by July 2026.

CAIS and Scale researchers' Remote Labor Index shows frontier models' success on real-world freelance tasks (3D/CAD, design, video, data analysis) jumping from 2.5% to 16.1% in nine months, with Fable 5 leading at 16.1%.

Center for AI Safety / Scale researchers · Import AIExtends · 1

DataArticle

Current AI agents remain far from reliable at long-horizon computer use, with the best configuration, Claude Opus 4.8 with maximum thinking, reaching only 20.6% binary accuracy on OSWORLD 2.0's multi-hour tasks, and performance dropping sharply as tasks lengthen.

OSWorld 2.0's creators find that even the strongest model setup only hits 20.6% binary accuracy on tasks averaging 1.6 hours of human effort, with agents struggling most on hidden-state recovery, tracking many items, and adapting to changing requirements.

OSWorld 2.0 researchers · Import AI

DataArticle

Current frontier AI agents remain far from reliable at long-horizon, multi-step computer-use tasks, with even the strongest configuration (Claude Opus 4.8) scoring only 20.6% binary accuracy on OSWorld 2.0, struggling especially with hidden-state recovery, tracking many items, and conflicting information.

OSWorld 2.0, a benchmark of 108 long-horizon computer-use tasks averaging 1.6 hours for a human to complete, shows even top models like Claude Opus 4.8 achieving only 20.6% binary accuracy, though rapid gains are expected as happened with OSWorld 1.0.

OSWorld 2.0 paper authors · Import AI