Link: LLM predictions for 2026 via Simon Willison

Based on his experiences exploring AI-assisted programming, Simon Willison shared his 1-, 3-, and 6-year predictions for LLMs on the Oxide and Friends podcast. In 2025, he found significant advances in coding agents and their ability to write good code. His predictions for 2026 focus on the problems and solutions for coding agent security, and his 3- to 6-year predictions focus on the changing role of software developers in building software.

Here are the predictions I shared in the episode.

• 1 year: It will become undeniable that LLMs write good code
• 1 year: We’re finally going to solve sandboxing
• 1 year: A “Challenger disaster” for coding agent security
• 1 year: Kākāpō parrots will have an outstanding breeding season
• 3 years: The coding agents Jevons paradox for software engineering will resolve, one way or the other
• 3 years: Someone will build a new browser using mainly AI-assisted coding and it won’t even be a surprise
• 6 years: Typing code by hand will go the way of punch cards

Via: https://simonwillison.net/2026/Jan/8/llm-predictions-for-2026/