Journalist in Paris publishing The French Tech Journal newsletter 🇺🇲🇫🇷. Launching: The France AI Radar: franceairadar.com/
Paris, France
- Tweets 48,845
- Following 14,630
- Followers 23,081
- Likes 179,567
Big Tech's greatest achievement may be turning the average American into a socialist...
🦔A farmer in Vineland, New Jersey photographed a neighboring golf course and accidentally caught 62 unpermitted gas generators at a data center that powers Microsoft's AI tools. Each ran at 1,982 kilowatts, over 50 times the state permit threshold. New Jersey fined operator DataOne $1.1 million, the largest fine ever against a data center in the state. DataOne can keep operating while it reapplies for the permits it already failed to get.
My Take
DataOne applied for permits, got rejected, and turned the generators on anyway next to farmland, homes, and two schools. The fine is a million dollars for a company with billions in capital. They decided it was cheaper to break the law and pay the penalty than to wait for approval. And they're probably right, because the fine doesn't shut anything down. They get 45 more days to reapply while the generators stay on.
Oracle can't get a gas pipeline permit in New Mexico. Over 500 communities have banned new data centers. And here's a company that just skipped the permits entirely and got a fine it can pay without noticing. The AI infrastructure buildout is on a collision course with every community it touches. These companies need power now and they're not going to wait for anyone to say yes. New Jersey already killed data center tax credits because residents organized. The more of these stories that come out, the harder it gets to site the next facility anywhere. The industry wants to spend a trillion dollars on infrastructure, but the people who live next to it have a vote too.
Hedgie🤗
2
600
Chris O'Brien retweeted
I am looking to hire PhD students, post-docs, and ML researchers in our Paris office interested in collaborating on fundamental research in AI model safety, not safety rules, not AI policy, not post hoc guardrails, not RLHF. We also have offices in Singapore, NYC, and Montreal. @amilabs #aisafety
37
106
1,226
116,560
Chris O'Brien retweeted
Meet my new company, North (@RunOnNorth): Claude, engineered into your law firm 🧭
6 months ago @ClarkValberg (@InVisionApp co-founder) called me out of the blue with "I have something we could build together."
It was law firms 👇
3
2
11
629
Chris O'Brien retweeted
After leaving @Dreamforce last week, I was inspired to revisit a question I hadn't fully addressed before: What happens when companies start building their processes and procedures into the weights of the models they use?
We're starting to see evidence of this with three recent product announcements:
@Salesforce’s Koa
@Harvey's Tenet
Recent @Shopify ML team experiments
With these specialist models, a software company can take a capable open-weight model, train it for a defined job, and lower the cost of delivering intelligence inside its product. Weights become another way to package and deploy expertise.
In that sense, weights are the new code.
And yet, if we play out this trend and specialist models become easier to build, lasting advantage may depend less on the weights themselves than on the process that improves them: the environment in which they practice, the judgment of what counts as success, and access to real outcomes with permission to learn from them.
So, to assess a company's position, one should ask: How much defined, repeated work can software companies take into their own hands? And does doing so create a lasting advantage?
I explore this in the latest edition of Decoding Discontinuity: decodingdiscontinuity.com/p/…
1
3
237
Uh-oh: The Brits ranked Paris as Europe's No. 1 AI hub.
Cue the hot takes served over a dish of cold baked beans on white toast, with a greasy fried egg, two sad pale sausages, black pudding, and a massacred grilled half-tomato.
observer.co.uk/news/technolo…
2
3
887
There are so so so many dumb takes on Mistral AI, alas, But if nothing else, at least ask: What does winning the frontier cost?
This is what it costs. Shoveling cash into a pit of fire. FOREVER.
And what do you get? Your ass eaten by Chinese models. And not in a good way.
2
5
622
This is not to say Mistral will succeed. Who knows? But I'd bet just as much on OpenAI being acquired by Bending Spoons next year.
1
1
88
I was curious about this because I have mixed feelings about @pangram. I had built a Pangram Claude skill via the API for other projects that provides a bit more detail.
But I ran Chapter 1 through and got a similar result:
Avez-vous entendu parler de ce roman ? C’est le « phénomène de la rentrée littéraire ». Il est lauréat du Prix Fnac, du Prix Méduse, du Prix Première Plume, fortement pressenti pour le Prix Renaudot, et il est en lice pour les prix Goncourt, Femina, Médicis et Décembre. Ah oui, il est aussi en cours de traduction dans 20 langues.
Eh bien selon Pangram, ce roman est rédigé presque intégralement par une Intelligence Artificielle. Nous avons testé plusieurs passages, à divers endroits du manuscrit, pour presque toujours le même résultat : 100% IA (cf les deux premiers chapitre dans le tweet ci-dessous).
Rappelons que Pangram est de loin le détecteur d’IA le plus fiable : son taux de faux positifs est de 0,0041%, soit environ un texte humain sur 24 400. Sur 996 273 textes antérieurs à 2022, Pangram en classe uniquement 14 (0,0014%) comme étant partiellement écrits par l’IA.
La probabilité que vingt passages d’un même livre (et il y a bien plus de 20 signalements dans le roman) soient identifiés à tort par Pangram comme étant produits par une IA s’élève donc à 0,0041% puissance 20, soit 0,1802 × 10⁻⁸⁵ %, c’est à-dire 0 suivi de 85 zéros après la virgule.
N’aurait-il pas été plus honnête que l’auteur crédite son co-auteur ?
3
5
1,312
Other big caveat: "If you know this text was written by a human, Pangram's explanation for false positives is stylistic convergence: a writer who has absorbed the cadence of generated prose, or who was heavily edited toward it."
Full breakdown is here:
frenchtechjournal.com/cetait…
122
The bros at a16z have some advice for Euro founders:
Think Smaller!
Noted...
sifted.eu/articles/a16z-andr…
4
518
The question in the end: is the book good or no? The rest is just Puratanism.
Avez-vous entendu parler de ce roman ? C’est le « phénomène de la rentrée littéraire ». Il est lauréat du Prix Fnac, du Prix Méduse, du Prix Première Plume, fortement pressenti pour le Prix Renaudot, et il est en lice pour les prix Goncourt, Femina, Médicis et Décembre. Ah oui, il est aussi en cours de traduction dans 20 langues.
Eh bien selon Pangram, ce roman est rédigé presque intégralement par une Intelligence Artificielle. Nous avons testé plusieurs passages, à divers endroits du manuscrit, pour presque toujours le même résultat : 100% IA (cf les deux premiers chapitre dans le tweet ci-dessous).
Rappelons que Pangram est de loin le détecteur d’IA le plus fiable : son taux de faux positifs est de 0,0041%, soit environ un texte humain sur 24 400. Sur 996 273 textes antérieurs à 2022, Pangram en classe uniquement 14 (0,0014%) comme étant partiellement écrits par l’IA.
La probabilité que vingt passages d’un même livre (et il y a bien plus de 20 signalements dans le roman) soient identifiés à tort par Pangram comme étant produits par une IA s’élève donc à 0,0041% puissance 20, soit 0,1802 × 10⁻⁸⁵ %, c’est à-dire 0 suivi de 85 zéros après la virgule.
N’aurait-il pas été plus honnête que l’auteur crédite son co-auteur ?
3
691
If I can't use Facebook agents to shop for SHEIN and Temu from Malaysian third-party hustlers on Amazon, is life even worth living?
6
589
The US is cooked...
🦔SoftBank is raising over $11 billion in junk bonds to fund its next OpenAI investment. One of the largest junk bond deals ever. The company has committed $65 billion to OpenAI total and is now the largest junk-rated borrower in bond markets in 2026. SoftBank's bond yield has climbed to 8.2% from 6.7% in January. The cost to insure its debt against default is at a three-year high.
My Take
Masayoshi Son bet $65 billion on OpenAI and is borrowing $11 billion more at junk rates to keep it going. He's borrowing at 8.2% to fund an investment in a company that isn't profitable and just pushed its IPO to 2027. The bond market is already pricing in the risk. The cost to insure SoftBank's debt against default is at a three-year high.
Son has been here before. He lost $70 billion in the dot-com crash because the future he funded didn't arrive on the timeline he paid for. He put billions into WeWork. The AI bet is the same structure at a larger scale. CoreWeave raised $3 billion in convertible notes last week. Oracle has $125 billion in liabilities. OpenAI spent $34 billion last year. Every dollar of this is borrowed at rates that assume AI demand holds at this pace for years. The Fed just raised rates and the 10-year is above 5%. Borrowing $11 billion at 8.2% to fund a company that burns cash, in this rate environment, is either the trade of the decade or the most expensive lesson Son has learned twice.
Hedgie🤗
1
681
A lot needs to be done to address liquidity in Europe. But obsessing about the US and IPOs is super weird. The IPO is an artifact. It's dead, except for a rare few IPOs, and most come to regret it...
Even Europe's best startups often end up legally incorporated in Delaware.
@SimonSchaefer, co-initiator of @euinc_petition, is pushing to fix that.
It means founders could go public without having to leave Europe first.
1
642
