ToolsThe story, in brief
This startup’s new mechanistic interpretability tool lets you debug LLMs
Goodfire just shipped a tool that lets you debug LLMs mid-training. Real-time model surgery is now possible.
The KeyNews take
Why it matters
Mechanistic interpretability moves from research curiosity to production tooling. Model builders can now adjust behavior in real time instead of retraining—this shifts the economics of fine-tuning and safety alignment.
The key facts
5 to knowGoodfire releases Silico tool for LLM debugging
Allows parameter adjustment during training
Enables fine-grained control over model behavior
San Francisco-based startup
Published April 30, 2026
Go to the source
MIT Technology Reviewtechnologyreview.com
Publisher excerpt: The San Francisco–based startup Goodfire just released a new tool, called Silico, that lets researchers and engineers peer inside an AI model and adjust its parameters—the settings that determine a model’s behavior—during training. This could give model makers more fine-grained control over how…