Claude Code made me love meetings again

The practice of software engineering is changing as we know it, that's clear to everyone who has used tools such as Claude Code or Codex. There were always people saying "Writing code was never the bottleneck", but I wholeheartedly disagree.

Just a few years ago (!), the only way to build an idea you had was to sit down and write the code. The code was definitely my bottleneck. You had to focus deeply, with the code executing in your head, the layers of abstractions all in your working memory. Whenever I was in that mode, I did not want to get interrupted or rather paused in what I was doing. I could certainly deal with a quick interruption, but anything that would destroy my flow? Hell no. Remember the times when you were debugging an issue intensely for an hour and a meeting came closer and closer like some sort of mental doomsday?

I started to loathe it whenever I was in such a state of deep flow / productivity and had to give it up for some meeting. This led me to continue to work on the problem if the meeting didn't require my full attention - and looking at the typing hands in those meetings, I sure wasn't the only one. When my mind is going 100 km/h solving a problem, I can't just force it to pull the handbrake to a stop and start thinking about a different one. This is also the reason why I had to stop working right before bed, as I couldn't sleep - my brain was still racing to solve the problem.

And that wasn't bad! As a productive software engineer, you wanted to be in that flow as much as possible. Only in that state could you deliver quality software at a good pace. I felt like I wasn't doing my job properly when I didn't manage to be in that state for most of my working hours.

But this is changing now, rapidly. I can have Claude Code or Codex write the majority of my code. I don't need to explore experiments myself anymore; it's often enough to give a rough outline and have it crank it out in 30 minutes.

The need to optimize my working time for these hyper-focused states of coding is gone. Of course I still have to do it, but it became much less than what it was. And I started to notice the effects of it: I started to schedule more quick meetings, and I'm more open to discussing things. Of course I was always "open" to it, but it was a delicate cost / price balance and I would prefer async communication always. But now... they feel more efficient. My mind is not burdened with deep logic most of the time; it's just high-level thoughts.

It will be interesting in which directions this will go - will we start using this mental capacity to run swarms of agents at the same time and spend our capacity on orchestrating them / working on multiple problems at the same time, as Steve Yegge suggests? Or will we become more "vertically integrated", meaning we own more parts of the product delivery - going more in the direction of a one-person company?

I'm more in the latter camp as of now; I think the increase in roles such as "Product engineers" is a good sign in that direction. Many of the new cohort of successful startups are hiring for that - be it PostHog, Ashby or Linear (to name a few).

Wherever it goes, for now I'm enjoying my freed-up mental capacity and increased capacity to get interrupted.