Taste and knowing what to build is the shit right now
Because AI tools can build anything, literally anything.
So knowing what to build and knowing what other humans need and want is key here.
Focus on that because everything else is quite easy to automate and will be even easier in the next few years
Ask yourself what other humans want and need, and NOT what you can build with 10 prompts in Claude Code, because that's not what you want to build.
You can literally build anything but you should build something other humans want.
My spiciest take: 🌶️
You only need prod.
Other environments are optional, and may cause more problems than they solve.
Imagine prod was our team’s only environment.
We’d:
- write lots of tests
- stop batching work in lower environments
- stop rushing to hit arbitrary release cutoff dates
- stop spending hours every week maintaining and coordinating work in lower environments
- auto-deploy upon merge
- release small changes multiple times a day
- monitor prod via automated checks that notify us of issues
- use feature flags and phased releases to safely test in prod before making a feature visible to everyone
These are mature dev team practices. Having only prod *forces* them.
So, your team might be better off with only prod than with a bunch of non-prod environments.
@thdxr I think it’s the same old engineering skills.
If you know how to break down a problem into small definable steps at the appropriate level of abstraction and complexity, agents will do just fine
Knowing what it is that you want remains the hard part
@sandislonjsak I mean, if understanding is fuzzy, clarity is off ofc. But even deep understanding often makes it hard to distill & articulate cleanly...
Agreed.
"It is about becoming clearer.
Clear on who you serve. Clear on what problem you solve. Clear on why you are credible. Clear on what you believe."