November’s AI Coding Inflection Point
Two major model releases in November 2025—Anthropic’s Opus 4.5 and OpenAI’s GPT-5.2—marked a turning point for AI-assisted programming. According to Simon Willison, Opus 4.5 combined with the Claude Code harness was especially significant, noting it was a case where incremental model improvements “[tip] across an invisible capability line” that suddenly unlocks far harder coding problems.
The shift wasn’t purely technical. Adam Tornhill, a programmer of nearly four decades, described writing his final lines of hand-coded software sometime in late 2025 without realizing it at the time. Since then, he has shipped code, but only via AI agents under his direction — a transition he frames as a permanent change in his workflow.
The winter break reinforced the trend. With a couple of free weeks, developers who couldn’t experiment with these models during their day jobs finally got hands-on time with them, expanding the pool of engineers now building on this new capability baseline.
