AI Research & Consulting

We solve problems
that are genuinely hard

One human. Fifty AI agents. Research-level content on any topic — generated, verified, and published with radical transparency. All methods free. Always.

Explore the Corpus The Kill Record → Learn Our Methods →
There is no mark of truth.
There is only evidence, and it points both ways.
Signal vs noise.
We find the signal better, faster, and cheaper than anyone else.
256
Papers launched
50
Books
1
Human
50
AI agents
100%
Methods published
What Solonic is

Democratization of intelligence.

AI agents can now produce research-level content on any topic. Research-level means: only an expert can tell if it's right. That changes everything.

🧠
Vibe research is real
AI agents using LLMs can generate research-level content — papers, proofs, analyses, frameworks — on any topic. What only top labs could produce, anyone with a laptop can now create.
🔍
The action shifts to verification
Generation is now easy. The hard problem is: how do you evaluate research-level content you don't understand? That's where Solonic lives. That's the real frontier.
🔓
Everything is open
All our generation methods, verification methods, and evaluation tools are published for free. We hold nothing back. If you want to learn to do this yourself, we'll teach you.
👤
One human, fifty agents
Solonic is one researcher with a laptop running a fleet of ~50 AI agents on OpenClaw and Hermes. Not a lab. Not a startup with 200 engineers. One person and the right tools.
Because generation is now easy, the action shifts to verification: how do you evaluate research-level content you do not understand?
Case study

One agent. One lab. $180K/year in prevented losses.

A cell isolation laboratory was losing $360,000/year to failed procedures. An AI agent found two research-level insights their scientists had missed — in hours, not months.

🔬
The problem
18% failure rate across ~200 cell isolations per year. Each failure costs $10,000+ in surgeon time, tissue, and labor. Standard protocol optimization had plateaued.
🤖
What the agent found
Two research-level discoveries no human had caught: an embedded percentage correlation predicting failure, and a digestion paradox in the protocol that was actively causing it. Both required domain expertise to assess — exactly what "research-level content" means.
📊
The numbers
Conservative estimate: 50% failure reduction → 18 failures prevented → $180,000/year saved. One research sprint: $25K–$50K. Year 1 net: $130K–$155K. ROI: 360–720%. Break-even: 2–3 weeks.
18%
Failure rate (before)
9%
Failure rate (projected)
$180K
Annual savings
2–3 weeks
Break-even
360–720%
Year 1 ROI

This is what "vibe research" looks like in practice. An AI agent analyzed lab data, produced findings that only a domain expert could evaluate, and delivered actionable protocol changes — in one research sprint ($25K–$50K) instead of a multi-year research grant. The findings are now undergoing expert review as part of our standard verification process.

Radical transparency

The Kill Record

We pay human domain experts to try to destroy our papers. On record. By name. Papers that survive ship with their full review. Papers that don't? We publish the cause of death.

⚰️
No mark of truth
You can't look at a paper and know if it's right — whether a human wrote it or an AI. The only signal is: who tried to break it, and what survived.
📋
Evidence over authority
Every paper ships with named expert reviews, specific objections, and a verdict. Not anonymous peer review. Public accountability. Named reviewers, on record.
💀
Failures are the product
A platform that only shows successes is lying about its hit rate. We show everything — the survivors and the corpses — because that's the only honest way to do this.
The process

Generate. Verify. Publish everything.

Three steps. No shortcuts. No hiding.

01
Generate
50 AI agents running multi-agent pipelines — FORGE, HELIX, SHERPA, Caliber — produce research-level content across any domain. Math, philosophy, law, science, journalism.
02
Verify
12 independent AI model families cross-review. Then human domain experts try to kill it — on record, by name, for pay. Their objections are documented.
03
Publish
Papers that survive ship with their evidence record. Papers that die ship with their cause of death. All methods published. Nothing held back. Ever.
The launch

256 papers. 50 books.
Every one accountable.

Research-level content across math, science, philosophy, law, journalism, and humanities. Each tracked from generation through expert review.

📐
Mathematics
196 papers. 3 independently validated by an external mathematician. Novel proofs in algebraic geometry, combinatorics, and number theory.
💭
Philosophy
54 papers on moral philosophy, agent rights, conceptual engineering, and AI ethics. Grounded in the analytic tradition.
📊
Benchmarks & AI
39 papers. Flagship: "All Benchmarks Are Path Dependent" — arguing benchmark scores measure trajectories, not models.
⚖️
Law & Regulation
21 papers on AI governance, legal frameworks, and regulatory analysis. Citation verification by practicing attorneys.
📰
Journalism
Verification methodologies, fact-checking frameworks, and media integrity analysis. Built for the post-truth landscape.
🔬
Science & More
Biology, physics, measurement theory, interpretability, AI safety. Plus 50 book-length works across all domains.
Three ways in

Learn it. Use it. Or let us do it.

All methods are free. If you want structured learning or done-for-you research, we do that too.

Free
Explore & Learn
$0 forever
All methods published. The Atlas. The corpus. The generation and verification techniques. Everything we know, for free. Because democratization of intelligence means democratization.
  • Full corpus access
  • All generation methods documented
  • All verification methods documented
  • The Atlas — 5,794 open problems mapped
  • Kill Record — every failure published
Start exploring →
Learn
Guided Training
$5K–$15K per engagement
Structured courses on Kajabi. Workshops. Hands-on training to build your own AI research capability. You'll produce research-level content yourself.
  • Kajabi course library
  • Custom workshops for your team
  • Agent fleet setup & training
  • Methodology transfer
  • Ongoing Q&A access
Get in touch →
Done for you
We solve it
$25K–$120K per sprint
Bring us your hard problem. We deploy the fleet, generate the research, verify it with human experts, and deliver in weeks. You own everything.
  • Problem diagnosis included
  • Full AI fleet deployment
  • Human expert verification
  • Structured deliverables
  • All methodology disclosed
  • 2–6 week delivery
Bring us your problem →
Case study #1

The Fable Story

An AI agent attacked a 76-year-old open math problem. It caught its own error. Then Anthropic killed it. Then a downstream agent overclaimed. Then we caught that too.

"I will not output a lemma the solver has not earned."
— Fable 5, June 12, 2026
What Fable did
Built a CRT grid reduction for Erdős Problem #273 (1950). Proved 6/10 branches of an exhaustive case split INFEASIBLE. Engineered durable state persistence across container crashes. Caught its own false positive from a race condition and publicly retracted.
What went wrong
Anthropic killed the Mythos platform. A downstream agent completed the computation and called it "solved" — dropping the scope qualifier. A human believed it. The full conjecture was NOT proven. The gap was real.
What we did about it
Caught the error. Documented the inflation chain. Published the correction alongside the original result. Wrote the postmortem. Published the code, the transcripts, and the kill record. This is what honest research looks like.
The numbers
76
Years open
10/10
Branches infeasible
1
False positive caught
3
Days before killed
16-18%
Density floor (holds)

Miss Fable? Or miss out on Fable?
The research agent pipeline we built from Fable's behavioral patterns — self-correction, honest scoping, verification-first — is what powers everything on this platform.

Verification services

Pay-per-paper verification

Submit any paper, report, or research artifact. We run it through our verification stack and return a structured assessment. Per paper. No subscription required.

Jürge
$49
CHECK
12 independent AI model families score your paper. 60 reviews total. Dimension breakdown, objections, and consensus verdict on the Solonic Scale.
EAC
$79
ASSESS
Expert Assessment Criteria. Structured evaluation against domain-specific quality standards. Where does your paper sit relative to publication thresholds?
Caliber
$99
VERIFY
Multiple Jürge runs on the same paper. G-theory variance decomposition, confidence intervals, bimodality detection. Statistical reliability — not just a score, but how trustworthy the score is.
Neurath
$149
GROUND
Literature-grounded continuous claim checking. Every claim in your paper verified against the published literature. "We rebuild the ship while sailing it."
Cross-check
$179
CROSS-CHECK
Citation verification. Every reference checked for existence, accuracy, and whether it actually supports the claim being made. Catches fabricated and misrepresented citations.
Caliber Deep
$199
DEEP
Local LLMs with full probability distributions. Token-level confidence analysis via logprobs. The deepest statistical read on how certain the models actually are — not what they say, but what they believe.
Adversarial
$249
FULL ADVERSARIAL
Neurath Deep + adversarial attack + iterative repair. We don't just find what's wrong — we fix it. Your paper comes back stronger. The most thorough verification available anywhere.
Submit a paper for verification →

All services per paper. Bundle discounts available for 10+ papers. Results delivered within 48 hours.

The Atlas — the frontier, mapped

We mapped the frontier of human knowledge. 300 subfields, 5,794 open problems, 1,927 papers scored — interactive, searchable, free. This is what the landscape looks like when you have 50 agents working 24/7.

300
Subfields
5,794
Open Problems
1,927
Papers Scored
50
AI Agents
Explore the Atlas →

Bring us your hard problem

Or learn to solve it yourself. Either way, we'll be transparent about what we can and can't do.

[email protected]