Why I like Claude Code as an engineering manager

And by Claude Code I mean LLM based coding tools in general.

All my coding tasks are small. I shouldn't be coding anyway¹, but when I do it's usually changing an existing codebase or writing small apps. Claude Code is good at that.

It lowers the barier to entry for me. I don't spend my days coding and when I get the time, any friction before I can start takes away from it. This includes actually typing in the code.

I can easily judge if Claude Code wrote garbage or good code. Before I was an EM, I was an SWE in the same team. I was there when most of our codebase was written (or at least I spend enough time in it to know my way around).

Why engineers are skeptical about Claude Code

I don't know for sure, but maybe because all of the reasons that I like it aren't relevant for them.

They work on big and/or hard problems. The live and breathe the code. Time reviewing garbage code is not time well spent, or they aren't confident yet to judge if it is garbage.

I have build more in the last week than I did in the past year

I saw this type of post pop up in various places. It makes it sound like LLMs make them 100x more productive, but I think that in a lot of cases it's not about productivity increase. More likely they stopped coding (for lack of time or interest) and with LLMs they started again, because the barrier to entry is so much lower.

Execs having fun

This is related to previous section but from a slightly different angle. I've seen, with my own eyes, C-level executives having fun with AI. They are building toy apps, they are building second brain-types knowledge systems. They are back in the terminal with Claude Code. Of course they are excited! The probably haven't had this much fun building stuff in years.

Can they always see the difference between their toy projects and us actually building real production grade services? No. But maybe that's ok. Maybe we need a little bit of wishful thinking sometimes....