How to decentralize a prover, according to an engineer who did it for fun

A software engineer has figured out how to decentralize provers in his spare time

article-image

Zac Durant/Unsplash modified by Blockworks

share

Andrew Milson, a former senior software engineer at Immutable, recently submitted his own proof to StarkWare’s StarkNet verifier on Ethereum.

Generally speaking, a prover is a computational entity responsible for attesting that information is accurate without revealing its underlying data. Provers must create “proofs” that can be easily verified by a verifier. 

The issue today is that many zero-knowledge (ZK) blockchain projects depend on one single, centralized prover. In the event that node fails to submit proofs, transaction finality will be impacted. 

But the process of decentralizing provers is no easy feat

Loading Tweet..

Milson’s interest in STARKs is what started the entire ordeal, he told Blockworks. 

“When I found out they could be used to prove very large computations instantly in a small proof, I just lost my mind. It didn’t even seem possible, and the mystery resulted in me becoming obsessed with figuring out how they worked,” Milson said.

During this time, he built miniSTARK, a tool that enables you to prove the integrity of arbitrary computations using STARKs and also spent time learning about Cairo — a STARK programming language. 

“Its simplicity is beautiful, and the fact that StarkNet is built on top of it strikes me as very strong,” Milson said.

Process of submitting proofs

STARK verifiers are very similar in structure to their provers, Milson said.

Being able to see StarkWare‘s verifier contracts on Ethereum essentially provided me with a blueprint that I was able to follow line by line,” he said.

Milson began copying Solidity contracts into a foundry project and wrote tests for the verifyProof process, he said.

It was an iterative process that involved the following:

  1. Modifying Sandstorm and miniSTARK
  2. Re-generating the proof
  3. Re-running the test

“Each iteration, I’d get slightly more lines of code passing until the whole test worked,” Milson said. “I used the passing proof and submitted it to Ethereum using Etherscanner. It was a real thrill once everything was submitted on-chain.”

Milson believes that being able to submit his own prover is a step towards decentralization for StarkNet. 

“Teams in the community can start integrating the prover into their work which will bring a lot of scale and interesting use cases to the ecosystem, for example, onchain Gaming and ZKML,” he said.

This sentiment is shared by Liron Hayman, the head of business development at Starkware.

“This is a very important milestone towards decentralization of the Starknet tech stack. Andrew was able to develop his prover without viewing the code of StarkWare’s prover,” Hayman told Blockworks. “A mature decentralized blockchain like Starknet requires a vibrant community developing multiple implementations of each component. Andrew’s work is taking STARK technology to serious decentralization — and by extension, towards fulfilling its potential.”

Milson believes his work can also be viewed as an audit of StarkNet’s existing verifier and a testament to the security of the network.

“I inspected every line of code and I couldn’t find a bug which was honestly a bit disappointing considering how juicy StarkNet’s bug bounty program is, but really impressive on StarkWare’s behalf,” Milson said.

Updated on Aug. 14, 2023 at 12:10 pm ET: Andrew Milson is no longer a senior software engineer at Immutable.


Get the news in your inbox. Explore Blockworks newsletters:

Tags

Decoding crypto and the markets. Daily, with Byron Gilliam.

Upcoming Events

Javits Center North | 445 11th Ave

Tues - Thurs, March 24 - 26, 2026

Blockworks’ Digital Asset Summit (DAS) will feature conversations between the builders, allocators, and legislators who will shape the trajectory of the digital asset ecosystem in the US and abroad.

recent research

Research Report Templates (5).png

Research

ERC 8004 introduces a new trust layer for AI agents by standardizing onchain identity, reputation, and validation. As agents begin handling capital and coordinating autonomously, trust becomes the key constraint to broader adoption. The rollout mirrors the early x402 narrative, where adoption lagged the initial launch until major integrations and a viral use case pulled attention into the ecosystem. If ERC 8004 follows a similar path, downstream infrastructure tied to the standard could see outsized benefit as the narrative gains traction. The primary beneficiaries are likely to be agent frameworks and launchpads at the distribution layer, agent to agent coordination platforms that enable delegation and payments, and validation providers that offer stronger security and execution guarantees.

article-image

BTC finished the week up 1.6%, while L2s, RWAs and the treasury trade continued to grind lower

article-image

DTCC moves DTC-custodied Treasuries onchain via Canton, while Lighter’s LIT launches trading at a fees multiple in Hyperliquid territory

article-image

In the 90s, rapt audiences worldwide watched a coffee pot — will that fascination ever turn to crypto?

article-image

Some systems improve by failing — and crypto has no choice

article-image

Yield Basis introduces an IL-free AMM design that already dominates BTC DEX liquidity

article-image

Maybe tokenholders don’t need the rights that corporate shareholders have come to expect

Newsletter

The Breakdown

Decoding crypto and the markets. Daily, with Byron Gilliam.

Blockworks Research

Unlock crypto's most powerful research platform.

Our research packs a punch and gives you actionable takeaways for each topic.

SubscribeGet in touch

Blockworks Inc.

133 W 19th St., New York, NY 10011

Blockworks Network

NewsPodcastsNewslettersEventsRoundtablesAnalytics