AI Made Coding Faster. Somehow the Workday Got Longer.

AI was supposed to remove some of the tedious parts of software development.

According to one engineer’s account, it did.

That may have created another problem.

An anonymous software engineer posting on X as @v0xium described joining a large company where Claude Code now generates much of the material surrounding software development: specifications, code, tests, product requirements, tickets, resolutions and reports.

The engineer says management has noticed the same productivity gains.

Their response, according to the post, has been to ask a straightforward question:

If pushing code is no longer the bottleneck, why are teams still moving slowly?

The post describes employees working 12- and 13-hour days while having increasingly little time to understand or review what the AI systems are producing. The author argues that engineers are becoming operators of an automated development process rather than participants in the underlying thinking.

The company is not identified, and Corporate Lunch cannot independently verify the employee’s description of its internal practices. The post nevertheless spread widely enough to draw public responses from prominent technology executives and was subsequently covered by Business Insider.

What makes the story interesting is not whether every detail of one anonymous workplace can be generalized to the entire software industry.

It is the management logic underneath it.

Productivity tools are usually sold with the promise of giving employees time back.

But companies do not necessarily have to give the time back.

They can also raise the baseline.

If a task that previously required a week now requires a day, that does not automatically create four days of breathing room. It may simply create an expectation that five times as much work should now fit into the same week.

That is where the productivity conversation becomes more complicated.

The constraint moves.

Writing code becomes faster, but reviewing it still takes time.

Generating specifications becomes easier, but understanding them still matters.

Creating tickets becomes nearly instantaneous, but somebody still has to decide whether the work described by those tickets is worth doing.

AI can dramatically increase the speed of one stage of a process without increasing the speed of the entire organization around it.

Management then faces a choice.

It can use the recovered capacity to reduce workload, improve quality or create room for deeper thinking.

Or it can update the throughput target.

According to @v0xium’s account, the second option appears to be winning.

The engineer’s complaint is therefore less about Claude Code than about what happens after Claude Code works.

The bottleneck disappeared.

Management noticed.

And instead of the workday becoming shorter, the expectation became larger.