Tech veteran Robert Cringely's startup 2Brains claims to solve AI hallucinations through a different architecture—separating language generation from fact-checking—rather than relying on scaling up model size.

What happened: Cringely, a 73-year-old tech industry observer, cofounded 2Brains Inc. in 2022 to address AI hallucinations (the tendency of AI systems to state falsehoods confidently). The company has patented an architecture that separates the part of the system that generates language from the part that retrieves and verifies facts, and reconciles the two before output reaches users.

Why it matters: The AI industry's prevailing consensus—articulated by influential figures—has been that hallucinations will solve themselves through scaling (making models larger and training them longer). Cringely argues this 'permission slip' has allowed companies to avoid directly tackling a problem he says is critical for enterprise use in hospitals, banks, and courts. His alternative approach achieves the result without waiting for prohibitively expensive scaling.

What to watch: 2Brains' solution runs on ordinary processors (consuming less power than specialized AI chips) and, on industry benchmarks for faithfulness, more than doubles the published baseline with no fabricated facts in the verified case. The company's core claim is that honesty and cost efficiency stem from the same design choice—looking facts up rather than guessing them.