Your own AI discovery lab.
Generative AI gives you answers someone already knows. Discoverative AI is built to find answers nobody knows yet — and we take on only problems that matter, and that reality can check.
Across AI research, software, scholarly work, biology and financial forecasting—and other verifiable problems we choose to take on.
Free to start.
For accepted projects, we build the lab and pay for the compute.
Pay only if it works.
If we hit the goal you set, you pay the price agreed in advance.
1,000 founding labs.
You pay nothing to start.
HOW IT WORKS
How it works.
Send us a problem.
Tell us what you want to find and how you’ll know it worked.
We build a lab for it.
We set up the tools, data and AI for your problem, and run the compute for you. Our research services are free within the agreed project scope.
Reality checks the answer.
Results are tested against the goal agreed at the start — by you, by an independent Reality Partner, or in Apodex’s own verification labs (opening soon).
If it works, you buy it.
Pay the price per the agreed terms. No royalties, no revenue share.
BEFORE WE START
Agree on Three Things
The goal
The result that counts as success.
The price
Starting from $250,000, set by what the answer is worth to you.
The deadline
Approximately 3–6 months, with check-ins along the way.
IF IT DOESN’T WORK
Check-ins along the way. At the deadline, you choose.
At agreed check-ins, you see the progress and the full lab record. Either side can adjust the plan or stop, as set out in the agreement.
If it worked
Pay the agreed price per the agreed terms.
Not yet
Agree a new deadline and keep going.
Stop
Walk away and pay nothing. What happens to the lab and any new results will be agreed upon before any work begins. Your own data and existing IP stay yours.
GOOD PROBLEMS
What problems are worth a lab?
Nobody knows the answer.
Not something you can look up — or ask a chatbot.
Reality can check it within ~3–6 months.
A test, a benchmark, a market, an experiment.
The answer is worth $250,000 or more.
Get it once and the answer’s value should far exceed the price.
OUR LABS
Eight labs.
One discovery system.
IN-HOUSE · RESEARCH PROGRAMME
AI Training Lab
Can AI run the full training loop itself? This Lab aims to use the latest authorized state-of-the-art models to independently build and train a model of at least 200 billion parameters—from data and experiment choices to training, evaluation and repair. The proof would be a trained model performing on held-out tasks, not an AI-written training plan.
IN-HOUSE · RESEARCH PROGRAMME
Complex Software Lab
Work toward autonomously developing complex software systems—from requirements and architecture to implementation, testing and repair. The proof is a working system that meets agreed acceptance criteria, not a convincing-looking code sample. End-to-end autonomy remains a capability to demonstrate.
IN-HOUSE · RESEARCH PILOT
Paper Review Lab
Papers are multiplying faster than conventional peer review can comfortably keep pace. A review from a single state-of-the-art model is not enough. We are testing a multi-perspective process that checks sources, methods, evidence gaps and alternative explanations, using TRACES to make the reasoning inspectable. Human reviewers and editors retain the final judgment; this is not an automated publication decision.
OPEN TO SELECTED EXTERNAL PROJECTS
Drug Research Lab
Help drug teams forecast clinical trial outcomes and investigate repurposing and reformulation. The goal is to turn evidence into a clearer decision: continue, adjust, gather more evidence or stop. The Drug Research Computing Center distinguishes tools available now from work requiring a research engagement.
IN-HOUSE · PERAXIS
Gene Delivery Lab
Design novel AAV capsids for gene delivery and test whether candidates meet defined delivery and targeting goals. A computational design remains a hypothesis until supported by experimental evidence; it is not a demonstrated therapy.
IN-HOUSE · PERAXIS
Weight and Metabolism Lab
Search for new weight-loss medicines beyond GLP-1 pathways. Explore alternative biological mechanisms and prioritize candidates for experimental testing. A research candidate is not a proven medicine.
IN-HOUSE · PERAXIS
Aging Prediction Lab
Predict a particular person’s future disease risks and health trajectory from their longitudinal health data—not just assign a generic biological-age score. Forecasts should state uncertainty and be checked against outcomes over time; they are not diagnoses.
IN-HOUSE · RESEARCH PROGRAMME
Financial Forecasting Lab
Turn financial and market questions into time-stamped probability forecasts, then check calibration and reasoning against resolved outcomes. On its first entry, Apodex-FutureFlow ranked #1 on a Prophet Arena weekly forecasting leaderboard (Brier score: 0.0897). That is a historical benchmark result—not a current-rank claim or a record of investment returns.
SHARE WITH US · INVITATION, NOT AN EXISTING LAB
What should we investigate next?
Tell us the question, relevant facts or data, how an answer could be tested, and why the result would matter. We review submissions and may take a problem into an existing lab or build a new one with you. Submission does not guarantee acceptance; scope, data access, validation and terms are agreed before work begins.
WHY US
Why us.
Our own models, end to end.
Your problem runs on models we built and operate ourselves. Your data is used only within the boundaries set in your agreement with us.
Built for discovery.
Our model was trained to find new answers, not to chat. It watches every step of the search as it happens to help catch mistakes early.
AI runs the lab. People make the calls.
AI builds the lab and runs the search, so labs start fast and move fast. Real-world tests and key decisions are made by accountable people.
Plugged into reality.
The AI lab knows which answers need a real-world test, asks for the exact test, and brings the results from real world tests straight back in — including from our own verification labs, opening soon.
Private labs for sensitive data.
For the most sensitive projects, we can establish private labs that can run inside your own environment, with none of your data kept by us.
DATA & IP
Your data and IP.
You retain any rights you hold in the data and IP you provide.
Apodex uses your data to perform your project, and will not share it except with confidentiality-bound partners as needed for your project, or as required by law.
You agree to share validation data from your project with Apodex, including negative and inconclusive results, subject to agreed terms and applicable law. Private labs are excluded.
Apodex’s platform, models, methods and know-how, including improvements to them, remain with Apodex.
Rights to project-specific findings and outputs — whether or not the project succeeds — will be agreed before the project starts.
The parties will agree on exact terms before any work begins. Apodex may change or end this offer of free research services at any time (for all users or individual users), but will honor any project agreement already signed.
FAQ
The questions we get asked.
APPLY
1,000 founding labs. Free while places last.
We read every submission and choose problems that matter and that reality can check and verify.
Apodex — AI for discoveries that matter.