Dylan Patel says Anthropic and OpenAI took about 30% of this year's new compute, and have 40-50% of next year's already signed
In an August 25 interview, SemiAnalysis founder Dylan Patel said OpenAI and Anthropic went from roughly 2 gigawatts each at the start of 2026 to above 5 by year-end, absorbing about 30% of all compute added this year, with 40-50% of next year's already under contract.
OpenAI publishes first Jalapeno results, claiming up to 1.9x more work per watt than the systems it tested against
OpenAI released measured results for Jalapeno, its Broadcom-co-designed inference chip, reporting 1.5-1.9x more AI work per watt, 1.7-3.6x lower latency, and 2.1-4.1x higher performance on interactive workloads, with kernels its own model wrote.
A forensic investigation fingerprints the anonymous free coding model that 491,000 developers have sent 42 trillion tokens
An independent investigator identified the anonymous 'Ox Alpha' model on OpenCode's free gateway as a Z.ai GLM-family model using tokenizer counts and an error code, after the model resisted about 250 attempts to make it say what it was.
An audit finds two released models silently reading future tokens, and the bug makes their own scores look better
Researchers found that inspecting the attention mask missed all 192 injected causality faults in their tests while a two-forward-pass audit caught every one, and the same audit found real defects in the shipped Zamba2 and Nemotron-H models.
Microsoft's AutoSaddler treats the agent harness as code to be patched, and gains about ten points on three benchmarks
Microsoft researchers built a system that reads an agent's failure traces, writes structured patches to the harness around the model, and keeps only the patches that survive validation, improving three separate long-horizon benchmarks by 9 to 10 points.
A new benchmark of 1,140 real agent failures finds the best method identifies the decisive wrong step 13 percent of the time
LongRCA Bench collects 1,140 genuinely failed agent runs averaging 145 steps each, with human labels for which step actually caused the failure, and finds that the strongest existing method locates that step correctly only 13.2 percent of the time.
Apple's Mac Studio now holds 512GB of unified memory, which solves capacity and leaves speed exactly where it was
The M5 Ultra Mac Studio configures to 512GB of unified memory at 1.2TB/s, enough to load almost any open-weight model in existence, but its memory bandwidth still sets a hard ceiling on how fast those models can generate text.
Anthropic says run-rate revenue passed $30 billion, up from about $9 billion eight months earlier
Anthropic disclosed that its run-rate revenue surpassed $30 billion, more than triple the roughly $9 billion it reported at the end of 2025, alongside a multi-gigawatt TPU expansion with Google and Broadcom starting in 2027.
Amazon quietly put Mechanical Turk in maintenance mode, and the shutdown date going around is not in any AWS document
AWS documentation states that Mechanical Turk is closed to new customers with existing customers unaffected and no new features planned, but no AWS page confirms the September 30 shutdown date circulating in coverage.