You're Accountable! Not the AI!

Every other week someone announces that software engineers are finished. The model writes the code now, they say. Pack it up. I'm not worried, and here's the bit the doomers keep skipping over.

Accountability can't be outsourced to a model

When production falls over at 3am, the model doesn't get paged. You do. It has no licence to lose, no reputation on the line, no manager asking what went wrong, no job to be sacked from. It cannot be held responsible for anything, because there is nothing to hold.

Every system that matters has a human name attached to it — the person who said "yes, ship it". That signature is the product. The code is just the thing the signature is attached to.

Someone still has to decide

A model is brilliant at generating options. Engineering is choosing between them with context the model will never have: the budget, the deadline, the politics, the half-broken legacy system nobody documented, the customer who will churn if this one feature slips. Judgement under constraint is the job. It always was.

Typing was never the hard part

Writing the characters was always the easy 10%. The hard 90% is deciding what to build, why, for whom, and what happens when it breaks at the worst possible moment. AI makes the typing faster. It does nothing to make the accountability disappear — if anything it raises the stakes, because now you're responsible for more code that you read more quickly.

So what actually changes?

Your centre of gravity shifts from author to owner. You review more than you write. You spend less time on syntax and more on architecture, trade-offs, and verification. The skill that pays is no longer "can you produce the code" — it's "can you tell when the code is wrong, and are you willing to put your name on it".

The day a model can be sued, fired, or wake up at 3am genuinely caring whether the payment ran twice — start worrying then. Until that day, you're accountable. And that, precisely, is why you're employed.