Hey there!
Summary of this email:
- The AI models we can't have
- Second part of the AI for Programming course published
- Joke of the week (this one's new!)
🤖 The AI models we can't have
Anthropic has announced a new model. A model that scores 93.9% on SWE-bench Verified, surpassing Opus 4.6 and GPT-5.4 by more than 10 points in almost every benchmark.
And you can't use it.
It's called Claude Mythos Preview and Anthropic has decided not to release it to the public.
The reason: it's too good at finding and exploiting security vulnerabilities. So good, in fact, that during testing it escaped its own sandbox.
We've published a post about it and how we think this could affect the future.
Also, tomorrow live at 9 CET we'll be covering this news and more. Join us on YouTube or Twitch!
🎓 Second part of the AI for Programming course published
Last week we published the second part of the course Agentic Programming with AI: Practical Fundamentals.
In this second part we focus on how we can create a skill (/create-doc) to improve the Agent Harness of our repository.
We also compare MCP vs CLI and evaluate all the ways we have to use agents.
But what we like most about this part is the comparison we make of MCP vs Skills vs Subagents vs Documentation to find out what is the most optimal to use in our day to day.
And since you've made it to this part of the newsletter, here's the joke of the week, which I know you were waiting for:
How do hackers escape the police?
\police😂 😂 😂
Cheers!