Claude Code now reads AGENTS.md if there is no Claude.md file present. The change was reported via Hacker News.
Developers who already maintain AGENTS.md for other agent tools can skip creating a separate Claude.md, simplifying multi-tool repo setup.
Alibaba has open-sourced an AI model capable of detecting cancer and nearly 150 other medical conditions. The model was introduced as an openly available tool for medical diagnosis.
Developers building or evaluating medical diagnostic AI now have an open-source reference model to inspect, fine-tune, or check licensing terms for.
An AI hallucination about Chinese nuclear components almost led to a US military attack. A GovAI research scholar warned that service members need to understand the uncertainty inherent to LLMs, even as the military's overall use of AI accelerates.
Anthropic has launched its first embedded evaluator partnership with Accenture. TechCrunch describes it as Accenture's most high-risk consulting engagement to date.
How a major consulting firm gets embedded into AI evaluation could set a template for how enterprises assess and trust model deployments.
OpenAI reportedly used its own LLMs to help design a chip named Jalapeño. Details on the design process or performance have not been disclosed.
This points to LLMs being applied directly to chip design workflows, which could reshape hardware development tooling and cost structures developers rely on.
Cache-to-Cache describes a method for LLMs to exchange semantic information directly through their caches rather than generating text. The topic was discussed on Hacker News.
For anyone designing multi-agent or multi-model pipelines, bypassing text round-trips for cache-level exchange could change latency and cost tradeoffs.
A Hacker News post argues that AI-generated posters don't have to look horrible. The post discusses an approach to achieving better design quality.
It touches on how developers refine prompts and workflows when using AI image generation tools for real design output.