Skip to content

Comment on Ask HN: Any Software Engineers here who enjoy their AI-native dev workflow?

Comments

For sure! I offload most work, except the "idea", rarely any domain-specific logic, and rarely any core design / architecture.

Mostly I treat the coding agent genie as a code calculator, building a plan, writing code, running tests, reviewing other code with scripts...

Fine to allow the genie to pick an idea from a queue of work, "what is most urgent and important" but not fine to go with whatever the genie thinks should be done - that's a fast track to building something you don't need yet.

I also run a pair-programming session once a week with my team where we direct the genie as a "third" person. It's important to me that we still socialize around the work, even if the genie is involved. Sitting around the pile of code on the table and riffing together is fun. Just because every team member has Claude doesn't mean we stop gathering around the table once a week.

Generally though, the enjoyment comes from the same problem-solving enjoyment I had when writing code before coding agents. It's fun to create things, fun to fix things.

The challenge is managing the addiction. I try to be done with any coding focused task by 2 pm, that let's me wind down and avoid any late into the evening wrap up. I'll spend the final hours of the workday reorganizing plans, reviewing PRs and admin tasks like email, IMs, etc..

Remember even though your enterprise is paying for your tokens, they are extracting 2x, 3x, 10x maybe even 20x the value from you with the heightened productivity from coding agents. If you have equity, let it rip ha but if you don't, take some of that productivity back for yourself.

take some of that productivity back for yourself.

Truer words hath not been spoken.

AboutSource Built by g1lg1l

Hackerly is an independent reader for Hacker News, built on the public HN API. Not affiliated with Y Combinator.