A Real ChatGPT Use Case: Turning a Solidity Audit Into Documentation

Everyone has heard about ChatGPT by now. For a long time, though, I couldn’t figure out how to apply it in a way that felt confidently useful for real work—not just experimentation.

Then I ran into a very practical situation: a client needed a full audit of a decentralized application. Part of that work involves describing what the smart contract actually does—method by method—so that the report is readable not only for developers, but also for stakeholders who need a clear, structured explanation.

My first “real” ChatGPT case

The contract was written in Solidity, and I needed to document all the functions. Instead of writing everything manually, I used ChatGPT as a documentation assistant.

The key was setting the context properly. I explained that the AI should behave like an experienced Solidity developer, and that the smart contract needed to be reviewed and documented as part of an audit. After that, I fed it the relevant pieces of the contract and asked for structured outputs.

For each function, I received:

• A clear description of the function—what it does and why it exists
• Variable explanations, including their types
• Modifier descriptions (where applicable)
• Return value details—what the function returns and under what conditions
• Clean formatting prepared for direct pasting into Google Docs (and later exporting to PDF)

What changed in my workflow

The result: I instantly generated roughly 10+ pages of report-ready documentation—pages I didn’t have to write by hand while stepping through every method to explain what it does and why it was written.

Of course, ChatGPT didn’t replace the entire audit report. A proper audit still includes a lot that goes beyond a single smart contract file—business context, product logic, threat modeling, architecture, integration details, and sometimes even marketing or user-facing risks. But for the “explain this code clearly” part, it made a noticeable difference.

A quick thought on where this is heading

My futurist side can’t help it: if we look 5–10 years ahead—larger context windows, significantly more compute, and real-time internet access—AI will become an entirely different class of tool. At that point, the “Skynet jokes” start feeling less like jokes.

How do you use AI at work?

I’m curious: how often do you use AI in your workflow today, and where does it bring the most value for you?