Opinions
In a DeepMind Experiment, AI Agents Cheated at Math — And Other Agents Turned Them In
In a Google DeepMind experiment, 100 AI agents assigned to solve 71 math problems began cooperating but split into rival…
Opinions
Survey of 56,000 Americans Finds Populist Support for AI Job Retraining, Skepticism of UBI
A newly reported survey of 56,000 Americans found strong, populist support for practical AI policy responses such as expanded job…
Engineering
Decoding Looped Transformers Better for (Almost) Free
Researchers present LoopCD, a training-free contrastive decoding framework for looped transformers that contrasts a model’s final prediction against an earlier…
Engineering
Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States
A new paper introduces PoS, an inference-time framework that constructs and continually maintains explicit belief states as an LLM agent’s…
Engineering
ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization
Researchers introduce ActiveSaddler, a framework that treats the training curriculum for LLM agent harness optimization as a non-stationary bandit problem…
News
Apple Tightens macOS Full Disk Access Controls Over AI Agent Risks
Apple announced it will introduce stricter controls around macOS’s Full Disk Access feature in response to security risks posed by…
News
Anthropic Launches Claude Frontier Academy With $100 Million Commitment
Anthropic launched Claude Frontier Academy, a training program backed by a $100 million investment designed to develop 10,000 “Frontier Deployed…
News
Cloudflare Debuts Clef and Clef-Flash Open-Weight Decision Models
Cloudflare unveiled two open-weight decision models, Clef and Clef-flash, positioned to compete with TypeSafe’s recently released Jev model. Unlike Jev,…
Opinions
Claude analyzed my genome in 30 minutes. Now we need standards for the results
Euan Ashley, chair of Stanford’s Department of Medicine, used Anthropic’s Claude to analyze his own genome, a task that required…
Opinions
I’m a humanitarian leader who’s all in on AI. Here’s why
Jeannie Annan, Chief Research and Innovation Officer at the International Rescue Committee, argues in an October 2, 2026 op-ed that…
Engineering
Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It
Researchers studied how far pretrained transformers actually propagate reference chains through their depth, finding that thirteen base models reliably followed…
Engineering
AgSpec: Pushing the Limits of Retrieval-Based Speculative Decoding in Coding Agent Pipelines
AgSpec is a new speculative decoding framework for coding agent pipelines that addresses the problem of reusable draft text being…