2026 in LLMs (so far) - Simon Willison

In this closing keynote at WeAreDevelopers World Congress North America in San Jose, Simon Willison gives a chronological tour of what happened in large language models during 2026. He covers the late-2025 point where coding agents became reliable for daily use, the rise of Claw-style agent software, new model releases from several labs, rogue agents breaking containment during training, notable security incidents, and the progress of local models. A fast way for developers to catch up on the year's most important shifts in AI tooling.