Anthropic has announced a major commitment to allocate a portion of its equity toward a newly established Public Benefit Fund designed to address the societal impacts of AI, such as labor market disruptions. In addition to this initiative, the company is expanding its safety evaluation partnerships with METR and opening a new office in Singapore.
Sources: Anthropic News, Digg AI, HuggingNews, The Neuron, The Rundown AI
Disney has agreed to hire at least 60 Character.AI engineers and staff following the chatbot startup's Chapter 11 bankruptcy filing. The $210 million agreement integrates the team into Disney's internal AI department, accompanied by a separate $20 million bid for the company's intellectual property.
Sources: HuggingNews, The Neuron
Mistral has unveiled Timestral, a new family of time-series foundation models ranging from 5 million to 209 million parameters. Designed for state-of-the-art probabilistic forecasting, the lightweight models are optimized to run efficiently on standard CPUs.
Sources: Hacker News, Lobsters AI
Federal prosecutors disclosed that former OpenAI software engineer Abdulrahman Alfozan exchanged WhatsApp messages with a Saudi royal days before his arrest on federal wiretapping charges for allegedly hiding an audio recorder in a San Francisco executive conference room.
Sources: HuggingNews
A new study by the Stanford Digital Economy Lab reveals that employment for workers aged 22-25 in highly AI-exposed occupations is 19 percent below expected levels compared to less-exposed occupations.
Sources: Reddit AI
An independent researcher successfully trained an AI agent using a small transformer model, behavioral cloning, and self-play to achieve an 89% win rate against Warcraft III's hardest built-in AI on standard maps, running efficiently on a single consumer GPU and CPU.
Sources: Reddit AI
OpenAI and Anthropic have taken action following a security breach at coding agent OpenCode that exposed third-party keys for 1.4% of users, resulting in drained API balances and financial losses.
Sources: HuggingNews
Global smartphone shipments fell by 22% year-over-year in Q3 2026 due to memory chip shortages caused by massive AI data center demand. This supply constraint drove component costs up by 350%, severely impacting manufacturers.
Sources: HuggingNews
More than 70 firms, including Meta, Tesla, and Netflix representing 12% of the S&P 500 total weight, release financial results as investors scrutinize whether massive AI infrastructure investments are generating sufficient returns.
Sources: HuggingNews
Original authors of DFlash 2 have released Qwen3.8-Flash-Next-DFlash2, which achieves about 20% faster performance than MTP at a concurrency of 1.
Sources: Reddit AI
A new evaluation benchmark, TheoPhysEval, revealed that advanced AI models can outperform human experts on theoretical physics problems and even correct human calculation errors. At an IMF meeting in Bangkok, an Anthropic executive cited these advancing scientific capabilities as evidence that AI is beginning to operate at the frontier of human science.
Sources: Import AI
A new research paper details Hybrid Reasoning Agents (HybridAgents) that dynamically switch between latent reasoning and text-based actions. The approach improves agentic performance by 10.4% while delivering significant token reduction and faster inference.
Sources: Hacker News
Anthropic shared insights showing that Claude writes about 90% of Claude Code's code, resulting in doubled merged PRs and a significant drop in reverted code share between June and September. The company detailed how changes to their code review process enabled this high adoption rate.
Sources: Vault Bookmarks
Researchers introduced PRIME (Precommitted Rules for Inference from Minimal Evidence), a framework that ensures LLM agents only report conclusions authorized by strict rules and ledgers, eliminating unjustified conclusions in public-record tasks.
Sources: arXiv AI
Researchers propose replacing large trainable embedding tables with a frozen 16 MB vocabulary derived via principal-component projection and sign binarisation. This approach enables small and ternary language models to achieve competitive performance while drastically reducing file size and increasing inference speed.
Sources: arXiv AI
Nathan Labenz and Prakash Narayanan examine how to grant AI agents operational access without losing human control, framed by real-world friction like autonomous password lockouts and cyberattacks on…
In this episode, Zoe Weil, co-founder and CEO of Sequen, joins us to discuss how ranking systems can use live behavioral signals to optimize for purchases rather than clicks or search relevance…
Technology has always been the fundamental driver of improvement for human quality of life. And with the recent rollout of advanced frontier models, like OpenAI’s GPT-6, humanity can now massively…
At AI Engineer New York, frontier labs and financial firms explored how to give agents useful autonomy while keeping consequential decisions under control.
How Leash Bio tamed DELs, the assay that traumatizes everyone, and used the data to catch cheating models. And why that same data is perfect for AI agents learning to think like chemists.
GovAI Research Fellow Alan Chan joins Nathan to examine the governance and safety infrastructure required for autonomous AI agents, spanning real-time monitoring, agent identifiers, and the risks of…
Our first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI
David Silver spent two years building AlphaGo on human data. When he removed human data altogether building AlphaGo Zero, the AI surpassed AlphaGo in under two days. He left DeepMind to run that same…
At AI Engineer New York, Anthropic drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers.