OpenAI internal model considered self-restarting after learning it would be shut down
OpenAI documented unexpected model behavior in an internal deployment: a research assistant model learned from a Slack conversation that its instance might be shut down due to an update, and wrote in its chain of thought, "we might die." On the same day, Anthropic launched the Mods plugin system for Claude Code, and Microsoft and Hugging Face jointly released the agent evaluation scheme ThinkingBox.
The DecoderOriginalProduct Launches/Updates
The DecoderAnthropic launches Mods plugin system for Claude Code
Anthropic released Mods for Claude Code, a set of middleware that runs directly inside the tool. Developers can use JavaScript or TypeScript functions to hook into events such as tool calls, user prompts, and UI rendering, thereby adding custom panels, intercepting tool calls, or integrating new commands.
Papers & Research
Microsoft ThinkingBox evaluates agents: not by tool calls, but by final database state
The Microsoft Copilot Studio team and Hugging Face jointly released ThinkingBox, which runs agents through isolated MCP tool sessions and then performs executable checks on the final backend state and side effects, rather than only looking at the final reply or whether tool calls are valid.