Delayed gratification is still one of the few bits of alpha remaining when using coding agents. Given how quickly they can churn out code, my heuristic is that I should spend ~75-90% of the time building something new on upfront planning and critical thinking.
Users affirm planning remains essential with AI coding agents because human oversight counters low-quality outputs despite advanced tools and features.
I really don't care what fancy skills, loops, orchestration, MCPs, blah blah blah that you're using.
That's all well and good. It's fun and can up your productivity.
But you are the bulwark against slop. Play your part!
Delayed gratification is still one of the few bits of alpha remaining when using coding agents. Given how quickly they can churn out code, my heuristic is that I should spend ~75-90% of the time building something new on upfront planning and critical thinking.
@marktenenholtz planning is where coding agents still need adult supervision
@marktenenholtz hmm the real grind is thinking before the agent types
most people skip this part and wonder why the output is junk
@marktenenholtz ngl most people skip the planning part entirely and just let the agent figure it out later
thats where the gap really shows
@marktenenholtz ngl i think this cuts deeper when you flip it the other way
how many hours wasted on bad code because someone skipped the thinking part?
Delayed gratification is still one of the few bits of alpha remaining when using coding agents. Given how quickly they can churn out code, my heuristic is that I should spend ~75-90% of the time building something new on upfront planning and critical thinking.