Jonathan Bouaziz

Deploying AI in the real world for 10 years.

[Jonathan Bouaziz @BouazizJonathan]

@LVMH_Presse

[Matthew Berman @MatthewBerman]

Google JUST announced it will invest up to an additional $40 billion in Anthropic and will provide Anthropic with at least 5 GW of computing power.

There were...hints...in my interview.

[Jonathan Bouaziz @BouazizJonathan]

OpenAI spent $100B teaching a machine to write. It still can't say 'I don't know.'

[Jonathan Bouaziz @BouazizJonathan]

We gave AI the entire internet and it still can't tell you if your email sounds passive-aggressive or just French.

[Jonathan Bouaziz @BouazizJonathan]

so here's your test. be honest w/ yourself.

1) can a well-funded team of experts replicate your core tech in 6 months? if yes, no technical moat.

2) if you lost 50% of your users tomorrow, would your product be meaningfully worse for the remaining 50%? if no, no network effect.

3) is every smart person telling you your market is too small, while you see a structural shift they're missing? if no, you're probably in a crowded market.

if you pass none of these, you don't have a company. you have a feature that's going to get absorbed by someone who does.

and imo that's fine to admit. better to know now than after 3 yrs and a series A.

[Jonathan Bouaziz @BouazizJonathan]

most startups founded in the next 5 yrs will die. not bc they're bad, bc they have zero moat.

here are the only 3 ways to actually survive 👇

[Jonathan Bouaziz @BouazizJonathan]

moat #3: the "too small" market.

Coinbase 2012. BTC total market cap was ~$1B. "internet money for libertarians and drug dealers." they IPO'd at $86B.

Stripe 2011. "PayPal already won online payments, why bother." lol.

Shopify 2006. "just sell on Amazon, why would anyone need their own store."

the pattern: everyone thinks the market is too small or too weird. but you see the structural shift that's about to make it 100x bigger. you're not betting on today's market. you're betting on a future most ppl think is unlikely.

this is contrarian in the real sense, not "i have a hot take on twitter" contrarian.

[Jonathan Bouaziz @BouazizJonathan]

ChatGPT doesn't make you smarter. It makes you faster at being wrong.

[Jonathan Bouaziz @BouazizJonathan]

Your AI startup is Openai's feature update.
At least don't share your codebase with them 😂

[Jonathan Bouaziz @BouazizJonathan]

the companies that will dominate AI in 5 years are mostly not AI companies. They're domain experts who figured out where models actually help.

[Jonathan Bouaziz @BouazizJonathan]

80% of AI startups funded in 2026 will be dead by 2028. Their entire product is something Claude can do with a well-written prompt. No moat. No proprietary tech. At least guys, build a great UI.

[Jonathan Bouaziz @BouazizJonathan]

Most AI companies in 2025: a Langchain script, 3 prompts chained together, a Stripe checkout page, and a pitch deck that says "proprietary AI."

[Jonathan Bouaziz @BouazizJonathan]

the person whose job AI is replacing will never tell you the AI works. they'll find every edge case, every bug, every reason it "can't be trusted." don't listen to opinions. look at the data. look at the big picture. that's where the truth is.

[Jonathan Bouaziz @BouazizJonathan]

The best AI products are the ones where users forget they're using AI.

[Jonathan Bouaziz @BouazizJonathan]

@summeryue0 I did it on my todolist, Openclaw had to rearrange my tasks and deleted the one "he thought" were late. It was late because I didn't do them on time mf. Next time, I'll respect my deadlines.