AI Didn't Write That Bug

“I keep fixing bugs that were written by the AI”

I hear that a lot, and I think it represents a breakdown of the social contract we have as engineers. Put very simply, it’s:

If your name is on the commit, then you did it

Do coding AI’s generate code with bugs? Yes, so do I and everyone else I know. It’s insane that we think anyone or anything can generate bugless code in a single pass. And even if it can, why would we trust it without testing?

It Reduces Your Value

Every time you say, “the AI did it”, it removes blame from yourself, but it also relinquishes claim to the value you bring to the team. When it comes time for budget cuts, it’s going to be harder to rationalize the value you bring.

It’s actually a very old problem. Every manager and higher-level IC has had to grapple with the idea that they’re no longer directly producing value, but instead have to work through others. An architect produces designs, but designs don’t run in production, so what value did they bring?

Accountability Is Key

Engineering managers don’t write code, typically. Yet they take responsibility for everything produced (or not) by their entire team. If they don’t take accountability for the good and bad alike, then they’re not a good manager. They can’t be effective.

The big societal shift is that, with AI, mid-level and junior engineers are having to take ownership of bigger pieces. The AI seems like an autonomous entity that’s able to solve problems. It takes wisdom and hard-earned soft skills to know when and where the AI can be trusted, when it needs to be validated, and when it should be ignored entirely.

Essentially, we’re throwing junior engineers into management roles without any help or support. We should at least recognize the trend, if we have any hope of addressing it.

A good starting point is: be responsible for the bugs you commit, regardless if an AI typed the code. Everyone needs to live by this.

Discussion