Accountability in the age of AI

Since more and more workload moves to LLMs, it makes some people think that perhaps some things are solved altogether. “Developers are cooked, it’s all automated”. But that assumes that all developers do is write code.

Now, I’m not saying there aren’t people in the world that do just that — people who are not able to do any work without a clear list of requirements. If any obstacle appears, they say “I’m blocked” without trying to even dig in or find solutions. No initative. Those people indeed may have issues compared to the AI since I’m not sure what they bring to the table in current times.

But I think (or perhaps I want to believe) that majority of the developers out there are interested in things they do at work. They are willing to go that extra step. Dig in further, propose technical solutions, be a partner in software creation and not a simple codemonkey.

What’s important for those people is also to be accountable.

It’s clear that we can use LLMs to write software at high speeds. Validation can be automated as well with tests/specs. Cursor or Anthropic already showed that codebases with strong test suites can be rewritten quickly with some success.

But when code is written fast and validated just as fast, can we deploy it at the same pace and stay accountable? Currently this may be a problematic step because taking full responsibility of what I’m deploying requires time and space. Shipping an update on Friday afternoon and leaving for the weekend doesn’t seem like a great idea after all.

It’s awesome that non-technical functions inside organizations can start delivering code. Designers can ship code. Product managers can test and prototype. Support can tweak confusing copy on pages. Marketing can build landing pages for their campaigns. But all of those updates had to be shipped. What if a silly copy tweak breaks the checkout page and returns 500s? Who will be on-call for that? I don’t think support or marketing is interested in that.

There are some orgs that fire people due to AI or put more pressure on people saying “with AI you can ship more!”. But at what cost?