← back

the fractal

> the year is 2025
> saas is my passion
> move to sf
> $100M revenue in 2 weeks
> I’m not a wrapper — I did post-training
> wake up, check X
> the labs have swallowed me whole
> back to zero

> let’s try b2b
> claude code: $20/month
> make something people want
> find a niche, launch on product hunt

> someone already built it
> we’re all writing the same code
> saas is a fractal
> saas is milk, eggs, air
> too cheap to meter

> all users churn
> time to sunset
> they were other b2b saas anyway
> b2b saas is a ponzi scheme

> read innovator’s dilemma
> okay — enterprise software
> blindside the incumbent with AI
> /claude write a salesforce clone with chat
> wait… how do these execs already know RAG?

> startups move fast
> incumbents have LLMs too
> senior engineers shipping 50k-line PRs
> /claude, give me a moat

> pick a vertical
> software eats the world
> I’ll sell to bankers
> 5x workflows (then replace them, shhh)

> find a banker, do the mum test, get pilot
> layoffs hit, they’re gone
> two weeks later — new job
> openai pays $500/hr for spreadsheets
> knowledge work: too cheap to meter

option 1: go to the party

> sf party
> researcher rushes over
> no, I don’t have dru— oh, you want training data?

> I’VE GOT IT
> sell data to the labs
> the real gold rush
> don’t sell picks and shovels
> just dig

> $100M ARR
> my data moves the species
> no craft, just arbitrage
> free money

> training data is eating the valley
> so agi can eat the world

option 2: look out the window

look outside
knowledge work: solved
but the future’s unevenly distributed
salvation lies in atoms
healthcare, construction, education — the frontier
far from text, for now

I pick construction
the built world is my passion