Skip to main content

The prompt as documentation: Should AI-generated code include its origin story?

In a recent code review, I stumbled upon something that made me pause: a developer had included the original AI prompt as a comment block above a set of classes. At first glance, it seemed like unnecessary clutter. But as I read through both the prompt and the resulting code, I realized I was maybe witnessing the birth of a new documentation practice that could fundamentally change how we understand and maintain AI-assisted codebases.

 


The case for prompts as living documentation

Everyone who took the time to dig a little deeper in using AI as part of his day-to-day coding activities knows that a carefully designed and written prompt can make all the difference. So, wouldn't it be unfortunate that all the effort we put into it is lost after the AI agent has done his job?

Some other reasons I could think of that makes storing these prompts valuable:

Reproducibility: If we need to modify the AI generated code, we could adjust the prompt and regenerate rather than hand-editing AI-generated code we might not fully understand.

Intent preservation: The prompt captures business context that might not be obvious from the code itself.

Debugging aid: When the code doesn't work as expected, the prompt reveals whether the issue was in the human request or the AI's interpretation.

The case against

But then I started to wonder; are we really going to start polluting our codebases with AI conversation history?

Noise vs. Signal: Not every function needs its origin story. A simple utility function generated from "write a function to capitalize first letters" doesn't warrant prompt documentation.

Maintenance burden: Who updates these comments when requirements change? Do we update the prompt, regenerate the code, then update the comment again?

Version chaos: What happens when we iterate on prompts? Do we keep a full history in comments?

The questions I can't answer (yet)

The more I dig into this, the more questions emerge:

Is this actually useful or just novel? Are we documenting prompts because they provide genuine value, or because AI-assisted development is so new that we're not sure what practices will prove valuable?

What's the half-life of prompt documentation? Code can last years or decades. Will a prompt from 2024 be meaningful to a developer in 2030 working with completely different AI tools?

Are we solving the wrong problem? Instead of documenting prompts, should we focus on making AI-generated code more self-documenting?

Where do we draw the line? If we document prompts for complex functions, what about any refactoring we do afterwards? 

Have you encountered prompt documentation in your own code reviews? What's your take—valuable practice or unnecessary overhead? I'm genuinely curious to hear from others wrestling with these same questions as we collectively figure out what good AI-assisted development looks like.

Labels:

Popular posts from this blog

Podman– Command execution failed with exit code 125

After updating WSL on one of the developer machines, Podman failed to work. When we took a look through Podman Desktop, we noticed that Podman had stopped running and returned the following error message: Error: Command execution failed with exit code 125 Here are the steps we tried to fix the issue: We started by running podman info to get some extra details on what could be wrong: >podman info OS: windows/amd64 provider: wsl version: 5.3.1 Cannot connect to Podman. Please verify your connection to the Linux system using `podman system connection list`, or try `podman machine init` and `podman machine start` to manage a new Linux VM Error: unable to connect to Podman socket: failed to connect: dial tcp 127.0.0.1:2655: connectex: No connection could be made because the target machine actively refused it. That makes sense as the podman VM was not running. Let’s check the VM: >podman machine list NAME         ...
Read more

Cache stampede: when our cache turned against us

While investigating some performance issues, we ran into an ASP.NET Core API that cached a fairly expensive aggregation query for 60 seconds. Under normal load, that was fine: one request rebuilds the cache, everyone else reads from it. Under peak load, dozens of requests would arrive in that same expiry window, all see a cache miss, and all fire the same expensive query in parallel. The database didn't like that. That was the moment when our caching layer stopped helping and started hurting. A burst of requests comes in at the same time, all miss the cache, and all go hammer the database or the downstream API at once. That's a cache stampede . The cache was supposed to protect our backend, and for a few hundred milliseconds it did the opposite. Why this happens IMemoryCache.GetOrCreate (and its async sibling) looks like it protects you, but it doesn't add any locking on its own. Look at the naive version: public async Task<Report> GetReportAsync(string key) ...
Read more

A complex system designed from scratch never works

A few years ago, I worked as an architect on a big mainframe rewrite. I still count it as one of my failures. Not because the technology was wrong, but because I couldn't convince the management team to simplify the approach. Years later, the organization is still struggling to get the new system up and running. I left the project at the time, because I couldn't put my name behind an approach that would take very long and cost a lot of money without a working system to show for it along the way. Gall’s Law That memory keeps coming back to me, because it's a textbook case of Gall's Law playing out in real life. Gall's Law , from John Gall's Systemantics , states it plainly: A complex system that works is invariably found to have evolved from a simple system that worked. A complex system designed from scratch never works, and it cannot be patched to make it work. You have to start over with a simple system that works. What does that mean in practice,...
Read more