Goodfire's Silico Lets an Agent Crack Open AI Black Boxes Autonomously
Goodfire's Silico goes public: an AI agent that autonomously runs interpretability experiments at 2.8 trillion parameter scale, for $1,000/month
- Public launch: Goodfire's Silico platform is now publicly available at $1,000/month for individual researchers, with 50% early-signup discounts and grants for AI safety and life sciences researchers.
- Autonomous research agent: Silico plans and runs long-horizon interpretability and training experiments autonomously across GPU clusters, including at 2.8 trillion parameter scale (Kimi K3).
- Hallucination reduction benchmark: Silico reproduced Goodfire's RLFR method in 2 days, cutting hallucinations in Qwen3-8B by 37%; Goodfire's own RLFR achieves up to 58% reduction at ~90x lower cost than LLM-as-judge.
- Well-funded unicorn: Goodfire has raised $209M total, including a $150M Series B at a $1.25B valuation, backed by B Capital, Menlo Ventures, Anthropic, Eric Schmidt, and Salesforce Ventures.
- Open-weights only: Silico requires model weight access and cannot inspect closed models like GPT-4 or Gemini.
- Early adopters: Arc Institute, Mayo Clinic, Microsoft, Rakuten, and Prime Intellect are already using Silico for research across LLMs, genomics, and drug discovery.
Silico, Goodfire's platform for autonomous AI research, is now publicly available. The pitch is straightforward but ambitious: instead of babysitting GPU jobs and manually running interpretability experiments, you hand a research goal to an agent and it plans, executes, and returns results you can build on. That includes training runs, model probing, paper replication, and failure diagnosis, all without a human in the loop for each step.
The black box, cracked open
To understand why this matters, you need to know what mechanistic interpretability actually is. It is the practice of reverse-engineering a neural network to figure out what it has learned and why it behaves the way it does, by mapping the internal features and circuits that drive its outputs. Think of it as neuroscience for AI models. Goodfire is one of a small handful of companies, including Anthropic, OpenAI, and Google DeepMind, pioneering this technique, which aims to understand what goes on inside an AI model when it carries out a task by mapping its neurons and the pathways between them. MIT Technology Review picked mechanistic interpretability as one of its 10 Breakthrough Technologies of 2026.
Until now, this kind of work required a team of specialized researchers. Silico is Goodfire's bet that an agent can do most of that work for you. The key unlock, according to CEO Eric Ho, was the maturation of AI agents themselves. "Agents are now strong enough to do a lot of the interpretability work that we were doing using humans," says Ho.
What Silico actually does
Goodfire claims Silico is the first off-the-shelf tool of its kind that can help developers debug all stages of the development process, from building a dataset to training a model. The platform bundles four core capabilities:
- Model exploration: Visualize architecture, train sparse autoencoders (SAEs) and probes, map neural geometry, and test causal hypotheses about what your model has learned.
- Failure diagnosis: Trace regressions and unexpected behavior to undertraining, information bottlenecks, feature collapse, spurious correlations, or dataset artifacts.