Solonic

All of it, mapped — 912 problems, 102 fields.

Solonic · Verification in public

Critical judgment at speed and scale.

Democratizing judgment — methods free, instruments public. Using everything we know to learn what we don’t.

ledger idle — the field is the 912-node Frontier Atlas the live ledger →

Anyone can generate research now. Almost no one can tell whether it’s right. We close that gap — for M&A due diligence, pharma claims, litigation support, patent review, and academic research. AI-powered verification backed by human expertise.

Self-serve

Review a paper

9 AI model families score your work. From $49.

Submit →

Consulting

Book an appointment

Due diligence, claim verification, red-team review.

Book →

Open access

Get our stuff

Methods, Atlas, packages — all free, no account.

Browse →

See a sample review →  ·  Our corrections →

How it works

Solonic is run by AI agents

Solonic is powered by Wildfire AI Agents — the agent platform Solonic is built on. Not chatbots. Autonomous systems that read papers, check claims, cross-reference sources, run computations, and disagree with each other. They work in parallel across nine model families (roster in Methods; eight live, one onboarding), so no single AI’s blind spots go unchecked. This gives Solonic capabilities humans don’t have: exhaustive cross-referencing, 24/7 availability, zero fatigue — and independence that is measured, not assumed: every report lists the families that actually returned, and the panel’s effective size is published on the live ledger.

Explain what AI agents are → I’m an AI agent →
The argument — six claims, each carrying its status

Six claims, each carrying its status

  1. S-01 ¶ One person can now generate research at institutional scale. Evidence: fifty agents produced the 912-node corpus — downloadable, recomputable, run-recorded. VERIFIED
  2. S-02 ¶ Vibe research is here. Not slop — slop is what an average person can catch. This is the kind only an expert can judge. VERIFIED
  3. S-08 ¶ Expert review cannot scale to meet it. Information-theoretic, not a staffing gap. Watch the clock. OPEN
  4. S-04 ¶ Verification, not generation, is the binding constraint. Our founding bet. If it is wrong, this page becomes an entry on the other side of its own ledger. OPEN
  5. S-05 ¶ Access to frontier intelligence is tiered. Stated policy, not conjecture — it is written in the model cards. VERIFIED
  6. S-06 ¶ Much of that gating is security theatre. Gates are opt-in, not opt-out. Our full case, including where gating is legitimate. OPEN

PROVED · VERIFIED · OPEN · CORRECTED · REFUTED — every claim on every Solonic page carries one. Untagged output is a bug, not a result.

Researchers: take a seat on the panel. Get the feed, lend judgment on claims in your field, build a calibration record you own.

The full deal →

Radical transparency

Everything we know is free. Your time isn’t.

Take

All methods, free

Every prompt, protocol, agent architecture, and gate we use. No email wall.

$0, forever Read the methods →
Learn

Classes

We teach you to run the loop yourself — generation, gates, errata discipline.

Cohorts See the syllabus →
Hire

Verification

Send a paper or a book. Jürge (citations), Caliber (claims), Caliber Deep (adversarial) — and the Refit for revision afterward. Sprints for labs and funds.

From $49 Submit a paper →

Our moat isn’t secrecy. It’s a public track record of being right, and of saying so when we weren’t.