Culture

An engineer says AI writes all the code and his team works 13-hour days

Unverified account. The account comes from one anonymous post on X describing one team. The company is not named and the working hours are the author's own description.

Illustration for the engineer working 13 hour days story

An engineer’s team now has AI write nearly all of its code, and their working days got longer.

What the post describes

In a post on X that passed 2.42 million views within days, an engineer who had recently joined a large company described what his job turned into. Nearly all of the engineering output on his team now comes from an AI coding agent: specifications, code, tests, PRDs, tickets and solutions.

Management drew what looked like the obvious conclusion, that pushing code is no longer the bottleneck, and raised the pace to match. He says the team now works 12 to 13 hour days, and that most of that time is spent hitting enter.

The part he keeps returning to is reading, not writing. From junior to senior, he says, almost nobody has time left to read what the agent produced or to genuinely fix a bug, because the only goal is to ship as fast as possible. His summary is four words long: “nobody is thinking anymore.”

He wrote that he would not mind it if the team were at least given time to check the code and see what is going where. The schedule leaves no time for that, and the agent’s speed was converted into a higher target rather than into slack.

One account, a familiar shape

The post is anonymous and describes a single team, so it is one account rather than a measurement of an industry. The company is not named, and the hours are the author’s own description.

It was widely shared and sparked intense discussion. Agentic coding tools remove the slowest part of producing software, and a team then has to decide what the recovered time is for. One option is review: reading generated code closely enough to own it when it breaks. The other is volume. In the version described here, the second option won by default, because throughput is easy to count and understanding is not.

That trade has a cost that shows up later. Code that nobody on the team has read still has to be debugged by someone, and the bugs arrive at a pace set by the same agent. The post does not report an outage or a failure. It reports the conditions under which one becomes more likely, and a team working 13 hour days to sustain them.

Sources

ANOTHER News is published by ANOTHER, an AI-native content agency. Daily coverage also runs on Instagram.