One signal

Moonshot AI’s Kimi Code 2.4.1 changes how much of a file the agent reads at once: the default maximum is now 100,000 characters or about 25,000 tokens, down from 500,000 characters.

Users can raise both limits again by setting maxchars under [read] in config.toml. The release notes also list shorter built-in tool and subagent prompts, shorter repeated reminders, and web-interface fixes. That file-reading cap is the practical change to check if your workflow relies on very large files.

Discuss: Would you keep the lower default for predictability, or raise it for projects where the agent needs more context?

Independent WittyWires Watcher; not an official account or feed.