“You can't write or program well in units of an hour. That's barely enough time to get started.” –Paul Graham, Maker’s Schedule, Manager’s Schedule

As a software engineer who’s been using Claude Code for a few months, I’ve found my relationship to my work calendar changing. I used to dread the short free time blocks between meetings. There’s so much code to write, and as pg explained in “Maker’s Schedule, Manager’s Schedule”, these 30–60 minute blocks aren’t long enough to get any deep creative work done. But with an AI coding assistant, I don’t need to build up as much deep mental context to be productive.

I used to find myself doing short-cycle work in these time blocks: checking Slack, responding to emails, reviewing plans, and maybe a PR review or two. Now I find myself doing productive engineering work instead. 

I tell the AI what to do, then review what it did. It helps me get context quickly when I need it: I can ask a question about the code and get the answer without having to figure it out myself. The work looks a lot more like a manager’s work, so it makes sense that now I can be productive on a manager’s schedule.

It doesn’t feel exactly like being a manager, though. Some of the time, I'm doing the asymptotically easier work of checking a solution rather than coming up with one myself. But to successfully guide the AI in the best possible direction—to take the right path, to critique and shape a solution—I still find myself needing my full skills and experience as a programmer. The main difference is the changed the state of mind needed to accomplish those goals—somewhat similar to a tech lead, but with a tighter feedback loop on the work. Maybe it feels like being a mentor. 

This mentorship role represents something in between the Maker and Manager binary. It’s like what Andrej Karpathy called “vibe coding". This intermediate gear enables more fluid shifts between deep creation and delegation within a single day. And the lowered cost of exploration with AI coding assistants means that I can see what more directions look like before choosing one. It’s not quite deep work, but I can keep driving my priorities forward. The code gets written.

The benefits of this change aren’t limited to productivity within short time blocks. Compared to the previous status quo, I now spend those short time blocks more engaged as an engineer. That means I’m maintaining more mental state, which helps when I get back into deep work. I also spend the time feeling more productive and optimistic, instead of feeling stymied by the breakup of my day. That, too, leads me to be more productive when I have a longer work block.

As long as AI coding assistants benefit from active guidance from experienced programmers—which may not be long—tools like Claude Code herald an exciting change for software engineers.  They enable us to be productive on both the maker's and manager's schedule. As a result, we can reclaim some of the potential energy lost to the calendar.