Evan Van Ness Calls Pearl 'Bigger Than Bitcoin' Over AI-Mining Hybrid

Evan Van Ness, a former Ethereum Foundation leader and venture fund manager, called the Pearl project "bigger than Bitcoin" in a November 11 interview with The Rollup, arguing it can run cryptocurrency mining on the same calculations that power AI inference.

Van Ness's claim rests on a concept he describes as useful proof of work. Bitcoin mining computations, he argued, secure Bitcoin but serve no external purpose, and researchers have spent decades trying to make proof of work perform socially useful calculations. Pearl, in his account, closes that gap by mining through matrix multiplication, the fundamental computation underlying AI and large language models.

How would the mining actually work?

Matrix multiplication appears in both AI training and inference, the process of generating responses to user queries. Van Ness said Pearl lets the same computation that supports mining also perform work AI services need. In effect, the hash-like security function and the commercial inference workload share a substrate.

The delivery mechanism is operational rather than architectural. Van Ness described Pearl's software as a plugin for VLLM, an open-source toolkit used to serve large language model inference. Operators could install Pearl alongside existing inference workloads and mine while processing those requests, without rebuilding their serving stack.

That framing matters for data center economics. If an inference provider can earn mining rewards on capacity it already runs for AI customers, the effective cost of serving model queries falls. Van Ness's argument implies mining becomes a byproduct of a revenue-generating service rather than a standalone, energy-intensive operation.

Does it improve efficiency?

Not according to Van Ness himself. Asked whether the approach made computing more efficient, he said it did not, explaining that adding noise to the calculations can introduce some inefficiency. The stated engineering goal is narrower: run mining alongside AI inference while keeping the additional computational burden as small as possible.

That burden varies by model. Van Ness claimed some models had shown no additional overhead across the full processing workflow, while conceding that skepticism about such claims was reasonable. Reducing overhead further, he said, remains an open engineering challenge — a caveat that separates Pearl's current state from the "bigger than Bitcoin" framing.

What is Van Ness's standing?

Van Ness said he discovered Pearl roughly 18 months before the interview and had spent more than a year absorbed in the project. His background spans a leadership role at the Ethereum Foundation and the management of a venture fund, which positions him as both an infrastructure insider and an investor commenting on a project he has followed closely rather than a neutral observer.

The comparison to Bitcoin sets an explicit benchmark. Bitcoin's proof of work secures a network that settled hundreds of billions in value; Pearl's wager is that piggybacking on the AI inference market — a market growing faster than any prior computing segment — can produce a comparably durable security budget at lower marginal cost.

The open question is verification. Van Ness's zero-overhead claims come from internal testing rather than independent benchmarks, and he invited skepticism himself. Until Pearl publishes model-by-model overhead data or third parties replicate results on VLLM deployments, the project's case rests on his account alone.