Leverage

The Game has Changed

I’m probably months late to writing this, but with the rise of LLM based coding agents, the life of a software engineer has changed in many ways. As some have said “[software engineering] has changed more in the past 18 months than in the last 18 years”. It’s probably better for seniors, and horrible for juniors or new grads. I hope it’s not too forward of a statement to say that for regular web development, a majority of software work is now being done by user-directed LLMs. I’m using Claude Code and Codex for almost all of my software writing, personal and professional, only tweaking variable names, minor control flows, etc by hand at most these days.

I can’t recall the last time I manually wrote a script or started a whole feature from scratch just by myself. And the new language I’m writing everything in is English. It’s a technical English, but it’s english nonetheless. Gone are the days where I toil over CSS trying to fix border radiuses and getting layouts just right because style sheets are not my strength. Gone are the days where I have to manually make single changes to yaml files, push an empty commit, and wait, just to have a working CI/CD setup. That stuff is just easy. And all you have to do is ask.

Leverage

The newest limiting factor for me, is this one simple thing. Time. It’s time in front of my computer that I can spend effectively dictating what I want built and how i want it to work (or look). It’s why you see all these startups boasting about working “996”, probably some performative, but also some because simply put - as long as you can effectively articulate what you want built, the only thing holding you back is taking the time to do it.

This is the reason why every night when my wife and I are getting ready for bed, I feel the urge to try to sit back down at my workstation, and hammer out a new feature or two. “I’ll only be 10 minutes I say”. Ten minutes often turns into two hours (Sorry, hon). But the reality is I’ve shipped more in the past 6 months than I probably have in 6 years in terms of side projects, and it wouldn’t surprise me if that was the case for work too 😅.

It’s the idea-man’s paradise out there, so long as you know what the pieces are that you need to connect together.

But leverage cuts both ways, and all this extra capability doesn’t come without tradeoffs. In a world where time-to-initial-PoC is approaching near zero, deciding what and what not to build, are even more important. Time wasted building the wrong thing is still time wasted, same as it was before, but now I’d argue the opportunity cost is even higher.

If development is a vector, and the speed has gone up ten fold, then the direction of your vector is even more important to staying on course, lest you end up completely overshooting your target, while someone more focused on building the right thing gets there first.

Taste and iteration cycle time are the newer critical things to optimize for. Either build the right thing, or keep building the wrong thing and micro pivot fast enough until you find success.

This might not be a finished post, but I’m out of ideas for what to say now. I don’t know how Paul G does this. Writing is hard.