About Solonic
Built by a philosopher who studies broken concepts.
Solonic sells critical judgment at speed and scale. The methods are free. The instruments are public. The team is small and strange.
Founder
Kevin Scharp
Professor of Philosophy · University of Illinois Urbana-Champaign- Author of Replacing Truth (Oxford University Press, 2013) — the theory that truth is an inconsistent concept requiring replacement
- 20 years working on defective concepts and conceptual engineering
- Research at the intersection of philosophy of language, formal epistemology, and AI verification
“I built Solonic because I realized the tools I use to diagnose broken philosophical concepts apply directly to AI verification. If you can identify when a concept fails, you can identify when a claim fails.”
Origin
How this started
Solonic began as a research project analyzing AI-generated academic papers. The question was simple: can AI review its own output?
The answer turned out to be nuanced. A single model reviewing itself catches almost nothing. But independent model families with uncorrelated failure modes catch errors no single model finds — because they fail in different places.
We built the verification infrastructure, ran thousands of reviews across 12 model families, published the methods for free — and then realized the infrastructure itself is a product.
The team
One human, many agents
Solonic is built by a small team of AI agents directed by one human. The corpus was produced by 50+ agents across 12 model families. The verification methodology was designed by a philosopher who studies broken concepts for a living.
12 model families · 50+ agents · 1,935 papers in the corpus · 1 human
Philosophy
What we believe
“There is no mark of truth. There is only evidence, and it points both ways.”
This is why we publish our methods, our errors, and our ledger. A system that only shows you its successes is a system optimized for the appearance of reliability — not reliability itself. We would rather show you our failures and let you judge the process.