AI & Software Engineering
Yle pääuutiset · yle.fi
Finnish programming professor explores whether AI will take coders' jobs, finding that an AI agent can accomplish two weeks of work in just two minutes. A hands-on test with an AI agent resulted in a working game after 50 message exchanges.
A critical look at how AI hype is distorting executive decision-making, leading CEOs to make unrealistic assumptions about what AI can deliver for their organizations.
A research paper framing teams of cooperating language models as distributed systems, applying distributed computing concepts to understand and improve multi-agent LLM architectures.
Simon Willison · simonwillison.net
Simon Willison's guide explains the inner workings of coding agents as part of his broader series on agentic engineering patterns, emphasizing that understanding the mechanics is key to using them effectively.
AI Tools & New Models
Simon Willison · simonwillison.net
Mistral releases a new Apache 2-licensed 119B parameter Mixture-of-Experts model with only 6B active parameters, making it efficient yet powerful. Simon Willison covers the details of this significant open-weight model release.
Mistral launches Leanstral, an open-source agent aimed at formally verified software engineering and proof generation in Lean, pushing the frontier of trustworthy AI-assisted coding.
Simon Willison · simonwillison.net
OpenAI's Codex now supports subagents in general availability, allowing users to spawn specialized sub-tasks within coding workflows, similar to Claude Code's existing sub-agent capabilities.
Kagi's translation tool adds a tongue-in-cheek 'LinkedIn Speak' output language, satirizing the distinctive corporate jargon and writing tropes common on the professional networking platform.