judgement per hour

everyone measures lines of code per hour

wrong metric

the one that compounds is judgement calls

---

most of the cycle isn't building

it's the dead air between

"this should work"

and finding out

---

tools collapsed that gap

idea → something exists → adjust

what's left is the decision

and the feedback

---

not building more

more reps on the read

closer to hitting balls

than reading about your swing

---

when iteration gets cheap

planning drops

just test

then test again

reads improve because volume goes up

---

cheap iteration has a cost

scope expands by default

---

the ones who pull ahead

won't be typing faster

they'll develop feel faster

most haven't noticed the shift yet

---

ai doesn't give judgement

that still comes from scars

it just removed the gap

between a thesis

and watching it break

---

real question

how many decisions got pressure tested today

that weren't testable a year ago

---

everything else is vanity