Noeri, founded by Stanford AI researchers Yejin Choi and Carlos Guestrin, publicly launched on November 16 with $110 million in funding led by Menlo Ventures. Its pitch: AI that learns a company’s work instead of treating the newest, biggest model as the answer to every problem. Source Source Source
Nvidia, Cisco Investments and Snowflake also invested, alongside Madrona, Kindred Ventures, Wing Venture Capital and NEA. The valuation wasn’t disclosed, according to GeekWire. Source
Self-improvement still needs a sign-off
According to the company’s description reported by GeekWire, Noeri studies business data and workflows, then proposes task-specific changes to instructions, tools, model selection and sometimes the models themselves. Each proposed change is explained and sent to a person for approval, much like a code review. The software gets to suggest the edits, not award itself an automatic yes. Source
On its website, Noeri says its software can learn from tasks, failures and corrections, refine or create tools, and train specialized models on a company’s work. Source
The company also says it works with closed and open-weight models. Customers retain ownership of the workflows and models they develop, subject to third-party model licenses, according to GeekWire’s account. Building something around another company’s model doesn’t make its license disappear. Source
A smaller-model bet, not a GPT-5 victory lap
Part of the idea grew from helping a Stanford microbiology and immunology team use OpenAI’s models more cheaply. Choi told Bloomberg that a small AI system using older OpenAI models outperformed GPT-5 for that team’s particular scientific research needs, an account also carried by GeekWire. Neither report supplies evaluation methods or measured costs for that comparison. Source Source
Choi frames the larger ambition as “a healthy ecosystem of many intelligences, in all sizes, strengths, and quirks.” Her launch post says Noeri aims to select the right-sized model for a problem at a fraction of the cost of frontier models. Noeri’s website says it measures quality and cost against customers’ own traffic and metrics rather than a public benchmark. Source Source
The cited launch materials and reporting don’t specify pricing or confirm general availability. They also provide no independent evaluation of Noeri’s performance or cost savings. Source Source Source Source
Comments