Asked what he would build if he started over, Patrick Collison did not name an AI lab. He said he'd go for "the janky 1990s software" running whole industries nobody in the Valley will touch. 3 examples of this: 1) Vertical SaaS

Aug 6, 2026 · 10:51 AM UTC

23
53
820
322,328
In many domains one dominant tool is left from the 1990s. Terrible mobile app, zero AI, awful collaboration. Collison's move is to build the good version for that one industry. His running joke: asked what tech he is excited about, drones or crypto or AI, he says vertical SaaS.
1
1
29
30,240
2) Domains software has never really touched The point is not a better app, it is a whole sector software skipped. On Replit, a founder built software for ice skating rinks, now headed for a multi-million dollar business. Collison: nobody in the Valley thinks about ice rinks.
2
67
31,613
3) What is popular with young people but low status The pattern is a market incumbents look down on, so no one serious is competing for it yet. Crypto was that ten years ago. Startups themselves were that fifteen years ago, which is why Stripe sold to them with no competition.
1
59
26,794
Why these three work now comes down to one shift. The time to reach a million, then ten, then a hundred in revenue has roughly halved versus the SaaS era. And 20% of new startups now charge a customer inside 30 days, up from 8% in 2020. The idea to income gap collapsed.
1
22
22,613
The domain expert is the one who wins here. Not the Valley programmer, but the person who already understands ice rinks, or classrooms, or a trade. On Replit a teacher who knew schools built MagicSchool, zero to half a billion dollars. He had the insight, AI gave the tools.
1
34
19,708
TLDR, where Collison would look if starting over: 1) Vertical SaaS, the good version of a cludgy 1990s incumbent 2) Domains software has never really touched, like ice rinks 3) What is popular with young people but still low status The domain expert wins, now that idea to income has roughly halved.
1
49
17,699
Sort replies: Relevant Recent Liked
Replying to @GeniusGTX
funny, because payments were janky 90s software until Stripe
1
10
2,455
Replying to @GeniusGTX
if i was starting over, i'd invent and patent scoop-shaped tortilla chips and sell to tostitos for a cool bill. instead i got stuck inventing frontier pq cryptography 😭
1
4,414
Replying to @GeniusGTX
Relius FIS and Prisim to name a few
1
2,384
Replying to @GeniusGTX
original interview link pls
230
Replying to @GeniusGTX
bs
167
Replying to @GeniusGTX
The boring, ugly software running banks, insurance, and utilities is genuinely a bigger and more defensible market than another AI wrapper, nobody's excited to rebuild it, which is exactly why the competition is thin.
1
4
1,598
Replying to @GeniusGTX
@grok which podcast is this video clip from?
2
2
3,578
Replying to @GeniusGTX
It’s true but we must mention that likely this software is being sold to SMEs. Known to be price sensitive and fairly fragmented. It’s why they are omitted by VCs. Takes much longer and growing it is like digging the ground for hours under the sun, like a farmer.
2
950
Replying to @GeniusGTX
Boring software running boring industries is still running the world.
1
726
Replying to @GeniusGTX
What ice skating ring software you referenced in the clip @patrickc @GeniusGTX
6,475
Replying to @GeniusGTX
AI transforming foundational, historically untouched industries like construction, energy, financial services, will be HUGE
176
Replying to @GeniusGTX
Built differently I guess
2,032
Replying to @GeniusGTX
Jajajajjjja he knows.
486
Replying to @GeniusGTX
Working on it 🫡
115
Replying to @GeniusGTX
Lots of opportunities
1,461