Independent Lean audit backs claimed proof of RH and Dirichlet GRH
An independent audit reports that the submitted Lean proof passes several kernel and statement checks for RH and Dirichlet GRH, with further checks still running.
Thinking Machines releases Inkling 2 Mini with reused transformer layers
Inkling 2 Mini reuses its middle layers three times and supports local inference on selected 32 GB systems.
METR reports faster growth in AI task horizons, with important measurement limits
METR estimates that its combined-suite 50% task horizon doubles roughly every 3.6 months, while warning of limits in human baselines and real-world transfer.
Anthropic closes IPO and $26 billion strategic private placement
Anthropic says its IPO generated about $91.6 billion in gross proceeds for the company, alongside a separate $26 billion private placement.
Researchers report a language model improving itself by editing its own weights
A model changed its own weight tensors without gradients and improved a held-out reasoning score in a controlled experiment.
Hugging Face trains coding-agent adapters for specific harnesses
Hugging Face reports that lightweight adapters trained inside coding-agent harnesses improved some matched software-task results, with mixed transfer to other workloads.