2,000 Submissions in Three Days: Opening Drug Discovery to Anyone Who Wants a Turn

Malaria kills a few hundred thousand people a year. Tuberculosis kills more. Neither is a mystery disease — we know the parasites, we know the bacilli, we have decades of literature. What we do not have is a pipeline of new drugs, because the people who need them cannot pay for the discovery.

That is not a science problem. It is an incentive problem. And an incentive problem is exactly the kind of thing you can route around if you make the work public.

What we opened

The Open Discovery Challenge is a drug-discovery competition that runs in the open on Hugging Face. Malaria and tuberculosis are among the targets — chosen precisely because medical need and commercial return point in opposite directions there.

The rules are boring on purpose:

• Anyone can submit candidate molecules. No institutional affiliation required.
• Submissions are scored automatically against the target.
• Results land on a public leaderboard that anybody can read.

No gatekeeping on who gets to try. If your molecule scores, it scores.

What happened

More than 2,000 submissions arrived within three days of launch. Two Korean outlets covered it on 18 August 2026 — Dong-A Ilbo and Electronic Times.

We did not expect that volume that quickly, and it is worth being precise about what the number does and does not mean. Two thousand submissions is evidence of participation, not of efficacy. Nobody has been cured of anything. A leaderboard score is a computational estimate, and the distance between a good score and a drug that works in a human body is measured in years and failures.

What the number does tell us is that a lot of people — inside labs and outside them — will work on neglected diseases the moment somebody removes the barrier to entry. The bottleneck was never willingness.

Why run it in the open

Closed benchmarks have a structural problem: you have to trust the party keeping score. Running in public does not eliminate that, but it changes what happens when the scorer is wrong. Anyone can inspect the ranking, submit a control, and argue with the result. Errors surface faster when more people can see them.

That matters more than usual here, because early-stage scoring is genuinely hard and every automated scorer carries assumptions that can mislead. We would rather find those in public than discover them three years downstream.

Take a turn

The challenge is live on Hugging Face. Submissions are open, the leaderboard is public, and you do not need permission to participate.

If you work on antimalarials or TB, or you just want to point a model at a problem that matters more than another chatbot benchmark — the door is open.

---

VIDRAFT builds sovereign AI models and evaluation infrastructure. The Open Discovery Challenge is part of our open-science work.