The best AI engineers think evolution not automation. Maybe it's there RL DNA.
The best founders do OODA loops, product improvement sprints, BML sprints, etc.
Your epistemological bottlenecks are always 10x more important than your material ones. UNLOCK GOOD DECISIONS FIRST.
I've been thinking about AI productivity incorrectly.
I focused on linear outputs: "how many worktrees can I handle," "how can my Mac Mini run all night" etc
But the biggest unlocks aren't linear...
You can even tell your AI, "Good decisions are my bottleneck. Help unblock me!!"
Once you make enough good decisions of the same type, you then automate (e.g., codify into a skill).
But the quality of your decision loop will always be upstream of automation.
Emerson has a quote on this:
"It is even true that there was less in [Napoleon and Caesar] on which they could reflect, than in another; as the virtue of a pipe is to be smooth and hollow."
That said I think "introspection" (spectere = to watch) is good; it's a means to hollow out. It's more "self-analysis" that's suspect.
@pmarca "It is even true that there was less in [Napoleon and Caesar] on which they could reflect than in another; as the virtue of a pipe is to be smooth and hollow." - Emerson
@aakashgupta Distribution has a supply problem too.
YouTubers are all talking about the "abundance issue." Ditto short form, ad creative, SEO, SDR automation, etc.
@gaganbiyani@mecolalu 💯 like going to the gym, everyone’s gotta get intellectually buff.
Meanwhile 10% self-learners can easily be half the market in consumer dollar terms.
@gaganbiyani Until recently it was hard to gamify fields with fuzzy answers and progressions and so a lot of edtech 1.0 (cohort/bootcamp aside) was either (a) gamified memorization or (b) passive consumption.