The other week I read Gergely Orosz’s excellent post on the state of the tech industry in 2026. In that post, several prominent developers share how they work with 5-10 parallel agents, each one running in a separate terminal tab. This is a trend that I think is about to enter the mainstream.

That trend also ties into a topic I’ve written about before: the cognitive cost of faster coding. Back then, I claimed that agentic coding is mentally expensive and more effortful than old-school manual coding. This time, I want to expand and revisit the human consequences of running five agents in parallel. I mean: what could possibly go wrong?

I’ve been sensing a real anxiety amongst developers over the past year or so around skill and knowledge atrophy. We know that AI tools like Cursor and Claude make developers faster – or at least feel faster – but at what cost?

Brian and his fellow researchers have now given that anxiety a name: the hollow orchestrator. A developer can easily find themselves approving more work each month, without becoming better equipped to judge it.

In this paper from Harvard, researchers look at a large dataset of approximately 300 million work events of 700,000 workers across 700 firms. They find that the adoption of AI agents leads to large and significant increases in lines of code (30%), commits (20%), and PRs (23%). These numbers are consistent with what I hear from talking to people in the industry, as well as in other literature. However, when they look at features deployed to production, they only find small (statistically insignificant) positive effects. How is it the case that so much more code is being written, but the amount of features we see varies so widely in different types of software?