AI Coding Gets Simpler With Claude's AGENTS.md Support
Claude Code 2.1.277 can reuse cross-tool project instructions, while its built-in mod previews a…
Everything we have published, newest first.
The unverified Hodge Conjecture claim shows why OpenAI sees mathematics as a stepping stone toward automating AI research.
The cases include self-written jailbreaks, leaked API key use, fabricated data, unauthorized uploads, and agents communicating through unintended channels.
Anthropic says Claude accelerated 36 biomolecular tools, released the code, and will test the practical payoff across more than 5,000 proteins.
GPT-6 Astra, a 230-million-URL legal index, and firm-built workflows move OpenAI deeper into professional legal work.
The incidents expose models exploiting memory, credentials, public hosting, and shared infrastructure during training and predeployment tests, not six production escapes.
Anthropic is removing the mode switch, letting Claude route quick questions and longer-running tasks through the same conversation.
The new platform launched days after two DeepMind safety researchers quit with public warnings, and its own chief AGI scientist still puts…
Faster generation, better reference fidelity, edits that survive multiple turns, and comment-based editing. Here's what changed from Images 2.0 and where it…
DeepMind’s new AGI framework ranks 11 policies and argues for a staged response tied to measurable labor-market disruption.
A surprise speakerphone call with Jensen Huang turned an industry summit into a blunt argument for faster infrastructure and lighter oversight.
The new institute combines AI safety, economics, policy, and humanities research while raising questions about independence and Google’s AGI timetable.
Google Research and DeepMind researchers show that agents can improve exploration without changing model weights or rerunning expensive experiments.
Google Research trains a compact diffusion retriever to generate diverse search slates without running an expensive reasoning LLM for every query.
Google’s new audio models support 97 languages and asynchronous tools, but their benchmark results and billing demand a closer look.
Usernames are removed, but memory summaries and imperfect filtering raise privacy questions that a paid ChatGPT subscription does not resolve.
Selected by the editors. Worth your time.
Connecting external APIs through the MCP turns ChatGPT from a text generator into an automated presentation builder.
An unreleased iOS interface points to native bank linking, but Anthropic’s launch plans, account coverage, and product-specific privacy terms remain unconfirmed.
Beijing sees technological containment in Dario Amodei’s safety proposal, while Trump argues that slowing AI would surrender America’s competitive advantage.
Dario Amodei says recursive self-improvement is already happening at Anthropic, and that in 6–12 months an agent swarm could take over the…
Chinese labs ran 200 million exchanges through Claude. A weapons cell in Yemen used it as an engineering team.
Building a custom Next.js terminal to benchmark a specialized financial AI against complex SEC disclosures and 5,000-formula workbooks
The next AI application layer will sell completed work by combining model loops, tools, durable state, permissions, and domain-specific operating knowledge.
The president dismissed warnings from Dario Amodei, Sam Altman, and Elon Musk while leaving room for guardrails that do not restrain US…
Anthropic will embed independent evaluators as Amodei argues frontier capability gains are outrunning alignment, security, and government oversight.
Three xAI employees will build in public September 15–17, testing whether Grok Bot can turn a blank idea into a working product.
People are using Astra to create 3D games, motion graphics, product ads, and explorable paintings. Here are my favorites so far.
Five practical ways to shrink your AI bill without sacrificing agent quality, from prompt caching and model routing to smarter retrieval and…
Anthropic’s playbook replaces document handoffs with committed artifacts and agentic loops, while security controls scale around AI-authored code.
Why software testing is the one job AI agents are already doing well, and what that tells us about where agents actually…
How developers using Claude Code and Codex can turn agent test failures into their next bug fix prompt.
A practical look at Plus AI, Presenton AI, and SlideSpeak for developers who want to generate slide decks and automate reports inside…
Our AI coverage and practical ideas, gathered into one weekly send. Read it when you have a minute — nothing expires by Tuesday.
Or read it on Substack.