Probably has developed a "data science mech suit," a deterministic validator that checks LLM outputs against a database and rejects mismatches, with the goal of stopping misinformation from reaching users. Its first product is an analytics tool that returns rapid answers from complex datasets along with source links and an audit history. The company emphasizes using much smaller, more efficient models—about four times less powerful than frontier models—that can run on local machines to cut AI token and infrastructure costs. Probably aims for 99.99% accuracy, comparable to traditional software, and believes clarifying context lets lighter models perform reliably. Founder Peter Elias has stated plans to expand the engine into accounting, financial reporting, medical diagnostics, legal document analysis, and technical engineering. The company recently raised $9 million from Andreessen Horowitz to further develop and deploy its technology.

Total raised

$9M

Funding rounds

1

Latest round

Seed

Latest activity

Jun 2026