Community Beat: Qwen 3.8, Claude Code in Bun/Rust, OpenAI Codex Context Cut (cont.)
: OpenAI reduced the Codex model's context window. Used to be three hundred seventy-two thousand tokens. Now it's two hundred seventy-two thousand. That's a hundred-thousand-token haircut. The community reaction was immediate — AmazingTurtle posted the PR, and the points tell the story. Two hundred sixty people thought this was worth flagging. Whether the reduction is about cost, performance, or something under the hood we can't see, it's a notable move for anyone building with Codex. So we've got a new model drop from Qwen, a runtime evolution for Claude Code, and a context reduction for Codex. Three different signals, same theme: the AI tooling landscape is moving while you're reading this. That's the AI beat. Coming up next — creative tools and gaming, because Blender dropped a major release and Minecraft went and switched its renderer.