When Thousands of Data Scientists Rewrite the Laws of Price Formation

What if 10,000 brilliant minds could tackle your company's most complex problems simultaneously? CrunchDAO has proven this isn't hypothetical—they've improved Harvard's cancer detection algorithms by 30% and are now bringing their collective intelligence protocol to Solana.

Summary

CrunchDAO, a decentralized platform connecting enterprises with a global collective of machine learning engineers, has launched its testnet on Solana, marking a pivotal moment in making advanced AI model development accessible to any business worldwide. Founded four years ago with a bold premise—that aggregating brilliant minds could solve problems no single organization could tackle alone—the platform has grown to encompass 10,000 ML engineers and has distributed over $1 million in direct cash payments to contributors.

Unlike typical AI startups building wrapper applications around existing language models, CrunchDAO positions itself as the solution for genuinely difficult, noisy, and complex modeling challenges. Their customers include sovereign wealth funds, major financial institutions, and elite research laboratories, all seeking to tap into collective intelligence that consistently outperforms internal benchmarks by significant margins.

The Solana testnet launch represents a fundamental shift in CrunchDAO's accessibility. Previously, only select enterprise clients could commission modeling markets. Now, any business sitting on valuable data can open a modeling market and harness thousands of data scientists to build custom solutions. The response has been immediate—300 applications received within two weeks of launch, with 77 teams already approved.

Key Points

The Collective Intelligence Model

CrunchDAO operates on a fundamentally different paradigm than traditional consulting or internal R&D. When a customer brings a problem to the platform, CrunchDAO opens what they call a "modeling market"—essentially a competitive environment where hundreds or thousands of ML engineers independently attack the same problem. The resulting models are then collected and aggregated into unified solutions that consistently outperform what any single team could develop.

This approach leverages the wisdom of crowds principle applied to machine learning. Different data scientists bring different perspectives, techniques, and expertise. When their models are combined intelligently, the ensemble often exceeds what even the best individual contributor could achieve. Over four years, CrunchDAO has collected approximately 40,000 models, creating a vast repository of approaches to complex problems.

Real-World Enterprise Impact

The platform's credibility stems from verified, substantial improvements in critical applications. One of the most impressive use cases involves the foreign exchange OTC market—literally the infrastructure for exchanging value globally. CrunchDAO opened a modeling market to develop pricing models, attracted hundreds of contributors, and now delivers 50-microsecond feeds to major banks that outperform competing Web2 solutions.

Perhaps most compelling is their work with the Eric and Wendy Schmidt Center at Harvard and MIT. This elite research lab was developing computer vision algorithms to detect cancer from cell images. After opening a modeling market on CrunchDAO, they saw a 30% improvement over their internal benchmarks. For a well-funded research institution that already employs top-tier scientists, such a leap represents extraordinary value.

The Abu Dhabi Investment Authority Partnership

CrunchDAO has maintained a three-year partnership with Adjalab, the research arm of one of the world's largest sovereign wealth funds. The sophistication of this collaboration is illustrated by the fact that they commissioned a Nobel Laureate to design the datasets used for modeling competitions. This isn't toy data or academic exercises—these are real-world problems where billions of dollars in investment decisions hang on model accuracy.

The improvements achieved for these clients aren't marginal optimizations. Jean Herelle, CrunchDAO's founder, emphasized they're achieving double-digit percentage improvements for organizations already investing millions into solving these specific challenges. This demonstrates the raw power of coordinated collective intelligence applied to finance's hardest problems.

Solana Testnet and Democratization

The recent testnet launch on Solana transforms CrunchDAO from a service provider into a permissionless protocol. Previously, engaging CrunchDAO required direct relationships and enterprise-scale budgets. Now, any organization can propose a modeling market, stake resources, and tap into the collective brain trust.

The choice of Solana as the foundation reflects practical requirements for this use case—high throughput, low latency, and low transaction costs are essential when coordinating thousands of participants submitting models and receiving payments. The protocol design allows for transparent incentive mechanisms where contributors are rewarded proportionally to their model's contribution to the ensemble.

Current Applications and Future Development

Beyond finance and healthcare, CrunchDAO is actively expanding into new domains. The Schmidt Center has opened a second modeling market focused on predicting the effects of DNA perturbations—fundamental research with implications for gene therapy and personalized medicine. This demonstrates the platform's versatility across radically different problem domains.

To accelerate ecosystem growth, CrunchDAO is offering five free spots for early coordinator teams. This initiative targets businesses with valuable data and meaningful problems who want to experiment with collective intelligence before committing significant resources. The goal is to prove the model's value across diverse industries and use cases.

Facts + Figures

• CrunchDAO has grown to 10,000 machine learning engineers worldwide, with 50% year-over-year growth

• Over $1 million in cash has been distributed to community members from actual customer revenue, not token emissions

• Approximately 40,000 models have been collected and catalogued over four years of operation

• The platform achieved a 30% improvement over Harvard/MIT's internal cancer detection benchmarks

• CrunchDAO's FX pricing engine delivers feeds at 50-microsecond latency to major banks

• The Solana testnet launched two weeks before this presentation

• 300 applications were received from teams wanting to launch modeling markets within the first two weeks

• 77 teams have already been approved as coordinators

• Five free spots are being offered to early-stage companies wanting to test the platform

• CrunchDAO has maintained a three-year partnership with one of the world's largest sovereign wealth funds

Top quotes

"If I can put together enough brilliant minds, and if I can focus them into solving some of the most complex problems in the world, I will be able to create a level of intelligence that nobody will ever be able to do."

"We are the guys that do the very hard stuff. We're not building LLM wrappers. We're not doing some fine tuning. We're building the models that nobody else wants to build."

"We're building the models that you'll find in the most noisy systems. The models you'll find for the most difficult and critical decision-making that some businesses have to make."

"They literally commissioned a Nobel Laureate to build the dataset that they put on the platform. This is really the level of complexity we are after."

"Guys, this is not nothing. We talk about one of the biggest research labs here. And we talk about major, major, major improvement."

"We have figured out over the last three years that if we can have hundreds of data scientists in front of enterprise problems, we can solve it faster. We can solve it better."

"For the first time, anyone on the planet, any business will be able to build and open a modeling market on Crunch."

"We can build your edge. We can build your alpha. We would probably make you your leader in your industry if you push data on the CrunchDAO."

Questions Answered

How does CrunchDAO differ from hiring traditional data scientists or consultants?

CrunchDAO operates through competitive modeling markets where hundreds or thousands of ML engineers simultaneously work on the same problem. Rather than relying on a single team's perspective, the platform aggregates diverse approaches into unified solutions that consistently outperform what any individual contributor could achieve alone. This collective intelligence model has proven especially powerful for complex, noisy problems where traditional approaches struggle to find signal.

What types of problems can CrunchDAO solve?

CrunchDAO specializes in the hardest modeling challenges—those involving noisy data, complex systems, and critical decision-making where accuracy matters enormously. Their track record includes financial market pricing, cancer detection from cell imagery, DNA perturbation prediction, and sovereign wealth fund investment modeling. If a problem can be framed as a prediction task with available training data, CrunchDAO's collective can likely tackle it.

How are contributors to CrunchDAO compensated?

Contributors receive direct cash payments based on how their models perform and contribute to the final ensemble solutions. Over four years, CrunchDAO has distributed more than $1 million to community members. Importantly, this money comes from actual customer revenue—businesses paying for models—not from token grants or emissions. This creates sustainable incentives aligned with delivering real value.

Why did CrunchDAO choose to build on Solana?

While not explicitly detailed in the presentation, Solana's high throughput, low latency, and minimal transaction costs make it ideal for a protocol coordinating thousands of participants submitting models and receiving frequent payments. The blockchain enables transparent, trustless coordination of incentives across a global community of data scientists who may never interact directly with each other.

How can my company access CrunchDAO's collective intelligence?

With the testnet launch, any business can now apply to become a coordinator and open a modeling market. Visit crunchtao.com/coordinator/proposal to submit your use case. If you have valuable data and a complex problem, CrunchDAO's 10,000 ML engineers may be able to dramatically improve your outcomes. Five free spots are currently available for early adopter companies wanting to test the platform.

What evidence exists that this approach actually works?

The most compelling proof comes from third-party validation. When Harvard and MIT's Schmidt Center—an elite research institution with world-class internal talent—saw their cancer detection benchmarks improve by 30% after opening a modeling market, it demonstrated CrunchDAO's value proposition. Similarly, major banks now rely on CrunchDAO's FX pricing feeds because they outperform traditional solutions.