The Moonshots Are In.

In early 2024, Andy Konwinski and I sat down to write a paper together. The question we were trying to answer was simple and uncomfortably large: how can computer scientists actually shape AI's impact on billions of lives?

That paper became the foundation for Moonshots, the flagship lab funding initiative of Laude Institute. The premise was straightforward: the most consequential AI researchers in the world should be the ones deciding how AI gets used, and they deserve the resources to think at the largest possible scale. We wanted to find out what would happen if we gave them permission to be wildly ambitious and held them to a real deadline.

We did not know they would respond. They responded.

125 proposals. 600 researchers. 47 institutions — virtually every top computer science department in the U.S. and Canada.

Fields Medalists, Nobel laureates, and Turing Award recipients pointing their best thinking at humanity's hardest problems. When we put out this call, we hoped for strong proposals. What we got was something else entirely.

The quality made selection genuinely difficult. Some of the most decorated scientists alive submitted proposals that did not make the cut.

// A word on the selection committee, because it deserves one:

In my fifty years of research, I have never been part of a committee like this one. We assembled research luminaries and rising stars from academia and industry — and crucially, world-class experts in each of the domains these projects are trying to move. John Jumper, whose AlphaFold work transformed structural biology. John Hennessy, Turing Award winner and former president of Stanford. Jeff Dean, architect of the computing systems that much of modern AI is built on. Eric Horvitz, Chief Scientific Officer of Microsoft. Yejin Choi, a MacArthur Fellow whose work on how machines reason — and fail to reason — has influenced a generation of AI researchers. Thomas Wolf, whose open-source work at Hugging Face put frontier models in the hands of millions of researchers. And domain experts of genuine stature — Mariano-Florentino Cuéllar on civic institutions, Erik Brynjolfsson on the economics of technology, Isaac Kohane on medicine, Ethan Mollick on how AI is actually changing work. Having that combination in the room shaped the standard against which every proposal was measured — and gave us the confidence to pass over proposals from Nobel and Turing laureates when we thought a less famous team had a more promising idea.

Here is what they are building:

Accelerating Science

Accelerating the Queen of Sciences — UCLA (Amit Sahai, Kai-Wei Chang, Raghu Meka, Nanyun Peng, Terence Tao, Wei Wang)

An AI system that conjectures, proves, and discovers mathematics the way a human mathematician does — learning, failing, speculating, and verifying, rather than pattern-matching its way to an answer. If it works, every quantitative science moves faster.

Actionable AI Weather Forecasts for Developing Economies — University of Chicago (Rebecca Willett, Ian Foster, Pedram Hassanzadeh, Michael Kremer)

Extending AI-driven weather forecasting to billions of people in data-poor regions. This team already reached 38 million farmers with improved monsoon predictions. The estimated value of better forecasts for Indian farmers alone: $3 billion.

Advancing Healthcare

The Virtual Embryo — Stanford University (Xiaojie Qiu, Emily Fox, Marinka Zitnik, James Zou)

The first predictive digital twin of mammalian embryonic development — a system that can simulate how an embryo evolves and identify where things go wrong. Congenital defects affect 1 in 33 births in the U.S.

JupyterHealth — Computational Precision Health / UC Berkeley / UCSF (Ida Sim, Ahmed Alaa, Irene Chen, Fernando Pérez, Maya Petersen)

Open artificial intelligence infrastructure that integrates wearable sensor data and electronic health records into continuous, real-time health intelligence. Built on Project Jupyter, the open-source platform already powering millions of scientific workflows, and extended now into medicine.

Workforce Reskilling

Collaborative Intelligence for the Future of Work — Stanford University (Erik Brynjolfsson, Tatsunori Hashimoto, Diyi Yang)

Granular measurement of how AI is reshaping 20+ job types, alongside collaboration infrastructure designed to augment rather than displace workers. Rigorous randomized controlled trials baked in from the start.

ARISTOS: Reskilling for a Physical Workforce — Carnegie Mellon University (Deva Ramanan, Changliu Liu, Raj Reddy, Katia Sycara, Jun-Yan Zhu)

An AI "virtual craft master" that teaches practical physical skills — drone repair, skilled trades — through real-time guidance and on-demand instructional video. The premise: dexterous physical work is among the last domains AI cannot automate. This project is the bridge for the workers who will need to get there.

Civic Discourse

Deliberation at National Scale — MIT / Harvard (Ariel Procaccia, Michiel Bakker, Bailey Flanigan, Archon Fung, Lawrence Lessig)

AI systems that allow millions of citizens to deliberate simultaneously in structured democratic discussion. Built on an existing, working platform. The five-year target: engage at least 0.5% of the U.S. population in genuine civic deliberation.

Rebuilding Trust in Civic Discourse — Cornell (Jon Kleinberg, Natalie Bazarova, Cristian Danescu-Niculescu-Mizil, Robert Kleinberg, Mor Naaman, David Rand)

A three-part AI trust infrastructure for online civic conversation — assistants that help individuals articulate views clearly, moderators that support group discussion, and an independent auditor that uses cryptographic verification to ensure the whole system stays neutral and inspectable from the outside.

// And then there are the 17 we could not walk away from.

We designated 4 runners-up and 13 honorable mentions across all four categories, and we are funding them too — $200,000 for each runner-up, $100,000 for each honorable mention, $2.1 million in additional grants in total. Later this year, all 17 will have the opportunity to pitch to philanthropic funders. That brings the full cohort to 25 teams now working on applying AI to what Andy likes to call "species-level" problems. The runners-up include a Stanford team converting 100,000 scientific papers into interactive AI agents, a University of Pittsburgh team building a perioperative digital twin trained on 1.5 million surgical cases, and teams from CMU and Johns Hopkins rethinking the relationship between AI governance and democratic institutions. The honorable mentions reach from Carnegie Mellon and MIT to University of Washington. I have seen enough of this field to know that Test of Time awards — selected a decade later — rarely go to the papers most celebrated at the moment of publication. We have high expectations for this entire cohort.

// One thing to look for.

If you click on the name of any researcher listed as a recipient, you will find something we have been building quietly at Laude: an Impact Profile. This is a first public look. The question underneath it is one we spend a lot of time on — how do you actually model research impact? We wanted to reach beyond paper citations and ask, bluntly: What did this person create, and what did that unlock for everyone else? We think the incentive structure matters enormously. When the best scientists in the world are genuinely optimizing for impact, the gains compound in ways that are hard to predict and impossible to capture in any standard metric. We are trying to build the infrastructure that makes that legible. More to come on this front.

// On how this gets funded — and why the model matters.

I have spent a lot of time thinking about what actually made the work possible over my career. RISC came out of a lab with a small team, a hard deadline, and a mandate to build something that worked — not just a paper that argued something might work. RAID, Spark, RISC-V: same structure. Interdisciplinary. Artifact-driven. Time-limited. Sunset clauses that forced urgency rather than drift. The returns on that model have been extraordinary — I recently documented over a trillion dollars in economic value traceable to $100 million in public investment across my labs. The model is not a secret. It just requires the right conditions to run.

Those conditions are under pressure right now. Federal funding is constrained. The private labs are luring the best researchers into closed, proprietary work. The open science ecosystem that produced the modern computing industry is being hollowed out from both ends.

Laude was built to model something different, and Moonshots is the first test of whether it can work at scale. We are not positioned against federal funding or against industry. We think there is a new model at the intersection of all three — academia, industry, and the public good — where philanthropic capital keeps the work open and pointed at outcomes that benefit everyone. Laude is anchored by Andy's $100 million pledge and funded by individual technologists who share that conviction. We have committed to fully funding at least one $10 million lab and are actively working with other funders to bring on additional support.

If this works — in the rigor of its process, the quality of what these teams produce, and the impact of the labs that come out of it — it becomes a path others can follow.

What Andy and I understood when we wrote that paper, and what this process confirmed, is that ambition was never the limiting factor. Six hundred researchers just told us exactly what they would do if someone gave them the resources and the mandate to think big. The answer was better than anything we could have dreamed. What has been missing is a funding model that matches their ambitions. Moonshots is our attempt to build one.