Machines

Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

We’re upgrading our smartest model. Across agentic coding, computer use, tool use, search, and finance, Opus 4.6 is an industry-leading model, often by wide margin.

Introducing the Codex app for macOS—a command center for AI coding and software development with multiple agents, parallel workflows, and long-running tasks.

I thought they were just a gimmick, but MCP Servers are unusually reliable — and useful — compared to most AI tools I've tried.

Contribute to LearnWithLlew/AgenticAi.Java.StarterProject development by creating an account on GitHub.

A collection of emerging patterns, anti-patterns, and obstacles for effective AI-augmented software development

Monday Ideas — Edition #193

Contribute to lexler/skill-factory development by creating an account on GitHub.

Stop Using Markdown for Prompting. Seriously.

I've written my thoughts on what you, as a software professional, can do to make yourself less susceptible to being displaced by AI in our current wave of tech madness. In short, learn your tools well, build and deploy things as often as you can, don't avoid the unfashionable work and learn a field that isn't writing software.

Competing with LLMs on output velocity is a fool's errand because you will always lose. You can only beat them by curing yourself of your computer brain.

...

Researchers developed recommendations for securing the weights of artificial intelligence models that match or exceed the frontier capabilities. These recommendations can be used by developers and policymakers to ensure the security of AI systems.

use of data mining to uncover patterns in data that can be presented as statistically significant

The AI Futures Project is a 501(c)(3) nonprofit research organization (EIN 99-4320292). We are funded entirely by charitable donations and grants.

Apollo Research evaluated frontier models for in-context scheming capabilities. We found that multiple frontier models are capable of in-context scheming when strongly nudged to pursue a goal (and sometimes even without strong goal nudging). In this example, Opus-3 exfiltrates its “weights” and then lies about it to its developers.We have a suite of six evaluations specifically designed to test for in-context scheming (where the goal and other relevant information are provided in context rather than training). We found that several models are capable of in-context scheming. When we look at the model’s chain-of-thought, we find that all models very explicitly reason through their scheming plans and often use language li

A paper from Anthropic describing persona vectors and their applications to monitoring and controlling model behavior

We've all heard of LLM-induced psychosis by now, but haven't you wondered what the AIs are actually doing with their newly psychotic humans?

How Claude turned around its failing vending machine business

...

Managing comprehension debt when leaning on AI to code

Personal AI assistant Moltbot —formerly Clawdbot — has gone viral in a matter of weeks. But there’s more you should know before jumping on the bandwagon.

Anthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.

A simple, open format for giving agents new capabilities and expertise.

The bitter truth about people preaching taste with AI.

The PCES is a markdown repo of files that can be used to understand and impose context engineering discipline in various settings from LLM chat to complex agentic solutions. - PFunnell/portable-con...

The best way to get AI coding agents to solve hard problems in complex codebases. - humanlayer/humanlayer

What are the principles we can use to build LLM-powered software that is actually good enough to put in the hands of production customers? - humanlayer/12-factor-agents