
I help people love their job. Product Management @aSparkFR , currently helping @betclic ship AI products. Proud investor in @lenarvermouth .
Paris/Villecresnes, France
Joined February 2009
- Tweets11.1K
- Following1.5K
- Followers1.5K
- Likes6.5K
Pinned Tweet
"Après six années passées à Istanbul, Nicolas Brun rentre au bercail pour assembler une carte postale sensorielle, une invitation au voyage aromatique frappée de soleil liquide"
whiskymag.fr/article/des-sco…
Benjamin Thomas retweeted
AI Strikes Again - Getty Images Heads For Chapter 11 Bankruptcy zerohedge.com/technology/get…
Benjamin Thomas retweeted
This OpenAI math drop is nuts.
This 7/8 quasi-Riemann hypothesis proof looks like one of the most important advances in analytical number theory in decades.
And they also resolved one of the most famous decidability problems in number theory!
Plus 3,998 more things! Oh my god.
Benjamin Thomas retweeted
Holy shit, America is cooked. They were able to perfectly clone the Microsoft Teams experience.
The European Commission has rolled out an alternative to Microsoft Teams.
But officials using it aren't impressed, with one describing the tool as “absolute shit.”
vist.ly/5kje9
Introducing Griffin, the first model to pass the video Turing test.
48% of people who talked to it live thought it was a real human. Previous systems have had a pass rate <3%. It is #1 on NVIDIA's benchmark for full-duplex AI video.
It’s the first Human Interaction Model (HIM).
Readers added context they thought people might want to know
The 48% figure and "video Turing test" claim are from Tavus's own study of 54 one-minute calls, not independently verified or using a standard protocol. Griffin-Lite leads NVIDIA's VideoFDB benchmark on their public leaderboard.
cellcog.ai/blog/tavus-gri…
research.nvidia.com/labs/amri/proj…
tech-ish.com/2026/10/02/tav…
Benjamin Thomas retweeted
looks like they solved it.
besides the fact that it reduced my head size by 20% at the 26th iteration and slowly made me fatter, everything else was almost intact.
my tattoos are also a bit faded on the 100th edit, but no one in their right mind should be doing that many edits.
very cool.
Benjamin Thomas retweeted
it's not a fad... it's not even a thing
literally zero serious F500 is post training their own model, no one wants to become the next BloombergGPT and have case studies written about how stupid you are in HBS for the next 5 years
Benjamin Thomas retweeted
John McCarthy (1927–2011) was an American computer scientist and mathematician widely known as the Father of Artificial Intelligence.
He coined the term “artificial intelligence” in the August 31, 1955 proposal for a summer research project at Dartmouth College. That proposal, written with Marvin Minsky, Nathaniel Rochester, and Claude Shannon, argued that “every aspect of learning or any other feature of intelligence can in principle be so precisely described that a machine can be made to simulate it.”
In 1956 he organized the Dartmouth Summer Research Project: the two-month workshop now considered the founding event of AI as an academic field. The gathering brought together the small group of researchers who would define the discipline for the next several decades.
Two years later, while at MIT, McCarthy invented Lisp (1958), a language built around list processing, recursion, and treating programs as data. Lisp became the dominant language of AI research for decades and remains influential. He also co-founded the MIT AI Project with Minsky and later established the Stanford Artificial Intelligence Laboratory (SAIL).
Beyond naming and organizing the field, he contributed time-sharing, garbage collection, situation calculus, and methods for representing common-sense knowledge. He received the Turing Award in 1971 and the U.S. National Medal of Science in 1990.
What can Astra do when given a humanoid embodiment?
We built HomeBody to find out. Controlled by GPT Astra, it carries out long-horizon tasks in a previously unseen kitchen—from tidying up across the room to retrieving remembered objects from ambiguous requests—without environment-specific training data or additional policy learning.
Here's how we did it 👀: tml.stanford.edu/homebody/
Benjamin Thomas retweeted
Opus 5.5 designing LEGO 👀
I asked it to design a Microduck I can build with real LEGO pieces. It:
> designed it life-size using 1113 real LEGO parts
> verified: 3,204 connections, 0 collisions, every step buildable, centre of mass inside the feet 🤯
> made a 141-page LEGO-style booklet (237 steps)
> priced every piece in the browser and prepared the orders on BrickLink
Enjeux des nouvelles techniques de Reasoning open.substack.com/pub/astral…
SITUATION DETECTED: A U.S. SOCOM analyst used AI to produce an intelligence report that hallucinated that a Chinese ship in the Middle East was carrying nuclear-weapons components. The U.S. military was preparing to board the ship before the error was caught, per CNN.
Benjamin Thomas retweeted
A mystery Polymarket user earned about $22,000 with a series of near-perfect bets on the quarterly earnings of companies that had one thing in common: they were all audited by KPMG. wsj.com/business/c-suite/clu…






























