Pinned Tweet
Scott Wu on the nihilism creeping into tech: people saying it's "too late" to build an independent company because the labs will just do everything.
His take: those people aren't founders. Founders are rationally optimistic with zero evidence they'll succeed.
If you think coding agents look like this forever, you're wrong. There are 10 more generations of this product to build.
Scott Wu: Easier Software Doesn't Mean Build Everything In House
"It doesn't actually change the math of what is your core competency that you really care about versus what are all the other things." — @ScottWu46 on @tbpn
2
22
AI Insights retweeted
Cognition's Scott Wu on why enterprise AI deals that normally take 12-18 months close in 3 for Devin:
- 25k engineers + $10B cost base = 3x speed is a priority, not a nice-to-have
- security reviews and private cloud deployment are the real bottleneck, not the tech
- they don't just hand over a tool, they forward-deploy and point teams to the exact use cases where Devin is 10x, and tell them honestly where it isn't
AI Insights retweeted
Scott Wu on why startups beat Google and Microsoft: you don't out-resource them, you out-care them.
Spotify's founder said it best when people asked why YouTube or Apple Music wouldn't just crush them: "we're just gonna care way more about music than they are."
Same logic for Cognition. Anyone can demo a sandboxed coding agent. Caring about the messy reality of shipping code inside Goldman Sachs or Mercedes is a different bet entirely.
Offer Steve Jobs $2 trillion to walk away from Apple. He says no every time.
Same reason Todd Graves keeps turning down absurd offers for Raising Cane's. It was never about the money.
It's about not being able to live with yourself if you didn't go for it all.
@ScottWu46 with @davidsenra on @FoundersPodcast
2
558
Scott Wu on AGI: no sudden singularity, just capability creeping up until "Devin ordering your DoorDash" quietly counts.
Then the Windsurf story: rumors Thursday night, cold outreach that evening, and a team that knew if they didn't have an answer by Monday morning, the whole thing would melt away.
5
1,105
AI Insights retweeted
People said “it’s too late” when GPT-3.5 was announced. Since then, companies have been built that are now experiencing unprecedented growth.
It’s only too late if you can’t think beyond the status quo.
Scott Wu on the nihilism creeping into tech: people saying it's "too late" to build an independent company because the labs will just do everything.
His take: those people aren't founders. Founders are rationally optimistic with zero evidence they'll succeed.
If you think coding agents look like this forever, you're wrong. There are 10 more generations of this product to build.
AI Insights retweeted
this is it. 1000% it.
people who think something is too late are usually not founders. one could argue that book business was done and amazon appeared to change the game. blockbuster was the king when netflix was the cd delivering peasant til the empire was overthrown.
again and again when the non founders think the game is over, the founders appear and change the game.
it's an infinity war if you choose to believe it, and the game is always on.
Scott Wu on the nihilism creeping into tech: people saying it's "too late" to build an independent company because the labs will just do everything.
His take: those people aren't founders. Founders are rationally optimistic with zero evidence they'll succeed.
If you think coding agents look like this forever, you're wrong. There are 10 more generations of this product to build.
Scott Wu on why startups beat Google and Microsoft: you don't out-resource them, you out-care them.
Spotify's founder said it best when people asked why YouTube or Apple Music wouldn't just crush them: "we're just gonna care way more about music than they are."
Same logic for Cognition. Anyone can demo a sandboxed coding agent. Caring about the messy reality of shipping code inside Goldman Sachs or Mercedes is a different bet entirely.
Scott Wu on the first time Devin actually worked: it fixed a MongoDB setup error autonomously, no copy-pasting Stack Overflow, no hand-holding.
"I could not sleep that night."
Few months later: 0 to $500M in under a year.
2
159
AI Insights retweeted
Getting security and deployment sorted that fast is honestly impressive.
Cognition's Scott Wu on why enterprise AI deals that normally take 12-18 months close in 3 for Devin:
- 25k engineers + $10B cost base = 3x speed is a priority, not a nice-to-have
- security reviews and private cloud deployment are the real bottleneck, not the tech
- they don't just hand over a tool, they forward-deploy and point teams to the exact use cases where Devin is 10x, and tell them honestly where it isn't
AI Insights retweeted
👀
Cognition's Scott Wu on why enterprise AI deals that normally take 12-18 months close in 3 for Devin:
- 25k engineers + $10B cost base = 3x speed is a priority, not a nice-to-have
- security reviews and private cloud deployment are the real bottleneck, not the tech
- they don't just hand over a tool, they forward-deploy and point teams to the exact use cases where Devin is 10x, and tell them honestly where it isn't
Scott Wu on why you shouldn't have to fight an email formatter in 2026: just tell the agent what you want and let it handle execution so you can spend your time on what to actually build.
Cognition's Scott Wu on why enterprise AI deals that normally take 12-18 months close in 3 for Devin:
- 25k engineers + $10B cost base = 3x speed is a priority, not a nice-to-have
- security reviews and private cloud deployment are the real bottleneck, not the tech
- they don't just hand over a tool, they forward-deploy and point teams to the exact use cases where Devin is 10x, and tell them honestly where it isn't
AI task length has been doubling every few months. Seconds to minutes to hours.
Pattern matching says that trend slows down eventually. First principles says ask why it wouldn't keep going, straight to days, weeks, months of autonomous work per task.
Everyone with an agent that can go do a month of work and come back with the result is not an incremental change. It's a different world.
2
189
Cognition's Scott Wu on how they found Devin's first real PMF: not the hardest architecture problems, not the mundane scripted ones — the repetitive-but-not-automatable middle.
Case in point: a 50,000 file codebase needing the same 8 changes everywhere. Enter Nubank, their first big win.
35
4,937
"Microsoft already has GitHub Copilot, why would anyone use a startup" is the same argument people made about Datadog vs the clouds, Snowflake vs Google.
The world has millions of problems to solve. Whoever has the most resources only wins if there's one end state everyone's racing toward. There isn't.
Scott Wu on why incumbents don't kill startups
7
1,263
Devin's first real task was fixing a MongoDB setup, the kind of error-loop hell every dev knows: Google the error, try the fix, hit a new error, repeat.
They just told it to go handle the whole thing itself. It worked. Cognition's Scott Wu said he couldn't sleep that night.
Not because of that one fix, but because of what it implied: why shouldn't all software get built this way now
4
210
Scott Wu was so mad about losing a math competition at age 7 that he still remembers it vividly.
Now he runs a company the same way: tree search, calculating moves, playing out the decision tree to victory.
Some people just see the world as one long competition.
2
84
Scott Wu: intelligence is no longer the bottleneck for AI.
Google got silver at the Math Olympiad. OpenAI landed top 100 in competitive programming. Models can already solve the hard problems.
What's left is the unglamorous part: teaching them the specifics of your domain, your libraries, your errors. That's the real work now.
3
100





