Dedaub Bytecode Decompiler and TheDAO Security Grant

If you’ve decompiled bytecode on app.dedaub.com, you may have noticed two improvements this year: more accurate, readable Solidity-like output and a larger free allowance for AI-powered reconstruction. Thanks to a grant from TheDAO Security Fund, we’ve been able to improve the application’s overall quality.

It started in person, when we met Jake and Griff from TheDAO Security Fund during ETHDenver week. We briefly discussed the Dedaub decompiler as a public good: a free tool that anyone can use to understand contracts published without source code. That conversation led us to enter the project in a quadratic funding round later that spring. This post explains where the funding went, what changed in the tool, and the story of the fund that made it possible.

The Dedaub decompiler: how we decompile bytecode into readable code

The Dedaub decompiler transforms deployed EVM bytecode without published source code into clean, Solidity-like code rather than raw opcodes. While our decompiler’s standard output is excellent in most cases, some low-level constructs may still appear in the final result. This is where AI reconstruction comes in: AI enhances this process by improving variable names and structure, making the output read more like developer-written code. The decompiler does the heavy lifting, while the AI layer polishes the result, and that is what the grant funded.

Why a decompiler? A contract without verified source still runs. Traces tell you what it did; only its logic tells you what it can do. The decompiler recovers that logic from bytecode. More than 30% of Ethereum addresses carry no source code, and malicious code mostly lives among them. That is why you need a decompiler: it is the tool you reach for when a contract has no published source, and you need to know what it actually does before you can decide whether to trust it or when you want to get to the bottom of a hack

This is not a new direction for us. The engine underneath, Gigahorse, has been running continuously since September 2018, first as Contract Library and now inside the Dedaub app, where hundreds of people use it every day, with more than 12,000 registered users in total across nine EVM chains, including Robinhood Chain, which was added just recently. The usage pattern says a lot about why the tool matters: when a hack happens, security researchers show up to decompile the attack contract, and bots deployed without source code get read the same way. We have written about the underlying decompilation techniques before: Gigahorse, our declarative decompilation framework, and Elipmoc, which pushed that further. AI reconstruction is a newer layer on top of that same foundation, not a replacement.

What we did with the grant

The proposal was specific, so the spend is specific too. The grant paid for two things:

Better AI models behind reconstruction. The reconstruction step calls a commercial AI model, and we pay for every single run. The grant let us switch to the latest frontier models, which means more accurate variable names and structure on the contracts people actually paste in.

Expanded free access to AI reconstruction AI reconstruction has a direct cost, so we previously capped daily usage to stay within our budget. The grant allowed us to increase the free AI reconstruction allowance for every registered user.

In short, the grant allowed us to improve our AI reconstruction in both quality and user quotas.

What TheDAO Security Fund actually is

TheDAO Security Fund holds a pool of roughly 75,000 ETH, most of it staked so the yield can fund security work across the Ethereum ecosystem on an ongoing basis. A curator group that includes Vitalik Buterin, Griff Green, Taylor Monahan, and Jordi Baylina steers that money toward wallet security, smart contract protection, incident response, and protocol security, aligned with the Ethereum Foundation’s Trillion Dollar Security initiative. Round 1 was the first time the money went out the door.

How the round actually worked

Round 1 ran April 23 to May 14, 2026, as a quadratic funding round on Giveth. Quadratic funding matters here because it doesn’t just hand the biggest check to whoever has the biggest existing following. Giveth used a matching algorithm called COCM, which weights broad support from many independent donors more heavily than the same total amount coming from a tight, coordinated cluster. A project that earns small donations from a few hundred different people gets more matched than one that gets the same dollar total from three whales.

The numbers give a sense of scale. Donors put in $315,020 directly across 13,805 individual donations from 3,934 people, and that was matched by a 637.4 ETH pool, anchored by 500 ETH from TheDAO Security Fund itself, topped up by Wintermute's $200,000 contribution and support from other contributors, forming the largest quadratic funding matching pool to date. Dedaub also provided each team member with $500-$1,000 to donate to projects they wanted to support, helping our team participate directly in the round.

Getting into the round was a filter in itself: 134 projects made it out of more than 250 applications, screened by Giveth and Ethereum Foundation reviewers against eligibility criteria covering smart contract security, wallet safety, monitoring, incident response, education, research, and standards. Once in, a handful of guardrails shaped where the match actually went: no single project could take more than 5% of the pool, every funded project was guaranteed a floor of at least 0.3339 ETH, and donations from verified ETHSecurity Badge holders counted four times as much in the matching formula, on the theory that people the ecosystem already recognizes as security practitioners are a better signal than an anonymous wallet.

Dedaub applied, passed screening, and raised $14,287.77 from 61 contributors, roughly $10,000 of it from the matching pool.

Thank you

Thank you to TheDAO Security Fund, to Wintermute for backing the round at the scale they did, and to Quantstamp, CredShields, and the 3,934 people who donated directly. We would not have shipped this upgrade this year without it.