Martin Signoux @MartinSignoux

Red-teaming innovation, fine-tuning regulation ——— AI policy @OpenAI perstechtives.substack.com Paris Joined June 2016

Tweets

3K
Followers

3K
Following

1K
Likes

13K

Martin Signoux @MartinSignoux

4 days ago

@matthiassamwald 👏 good to see you again in the AI Act space Matthias

0 0 1 137 0

View Details

Martin Signoux @MartinSignoux

4 days ago

@Manderljung 👏From GPAI Code drafting to implementation

0 0 1 83 0

View Details

Martin Signoux @MartinSignoux

4 days ago

@Miles_Brundage Good choice !

0 0 1 79 0

View Details

Noam Brown @polynoamial

a week ago

After AlphaGo, the skill of human Go players noticeably improved. I suspect we will see a similar pattern in math.

Timothy Gowers @wtgowers @wtgowers

a week ago

Another major problem, this time in additive combinatorics, has fallen, this time to humans rather than AI, but using methods related to the AI solution to the unit distance conjecture.

25 162 2K 798K 330

186 975 9K 775K 2K

View Details

Martin Signoux @MartinSignoux

2 weeks ago

@yonashav 🫡

0 0 0 171 0

View Details

Martin Signoux @MartinSignoux

2 weeks ago

@GrablyR Quel modèle avez-vous utilisé ?

0 0 1 433 0

View Details

Martin Signoux @MartinSignoux

3 weeks ago

@Fabien_Mikol x.com/OpenAI/status/… Les IA qui ne savent pas raisonner 😇

OpenAI @OpenAI

3 weeks ago

1K 4K 27K 13.5M 9K

0 0 3 408 0

View Details

Today, we share a breakthrough on the planar unit distance problem, a famous open question first posed by Paul Erdős in 1946. For nearly 80 years, mathematicians believed the best possible solutions looked roughly like square grids. An OpenAI model has now disproved that belief, discovering an entirely new family of constructions that performs better. This marks the first time AI has autonomously solved a prominent open problem central to a field of mathematics.

1K 4K 27K 13.5M 9K

View Details

Noam Brown @polynoamial

a month ago

After 100 million tokens, performance was still going up. What we're seeing here is not the capability ceiling. From the report: "Performance on TLO continues to scale with the amount of inference compute spent, and we have not yet observed a plateau with the best models."

AI Security Institute @AISecurityInst

a month ago

OpenAI’s GPT-5.5 is the second model to complete one of our multi-step cyber-attack simulations end-to-end 🧵

95 396 2K 1.8M 748

33 125 1K 198K 373

View Details

Noam Brown @polynoamial

a month ago

A hill that I will die on: with today's AI models, intelligence is a function of inference compute. Comparing models by a single number hasn't made sense since 2024. What matters is intelligence per token or per $. This is especially true when using it in a product like Codex.

Lisan al Gaib @scaling01

a month ago

The GPT-5.5 model family completely dominates the cost-performance frontier on the Artificial Analysis Index

34 39 874 248K 105

46 97 1K 128K 305

View Details

Martin Signoux @MartinSignoux

2 months ago

@benedictk__ 🧑‍🍳 Tasty !

0 0 0 37 0

View Details

Forecasting Research Institute @Research_FRI

2 months ago

We completed the most comprehensive study of how economists and AI experts think AI will affect the U.S. economy. They predict major AI progress—but no dramatic break from economic trends: GDP growth rates similar to today's and a moderate decline in labor force participation. However, when asked to consider what would happen in a world with extremely rapid progress in AI capabilities by 2030, they predict significant economic impacts by 2050: • Annualized GDP growth of 3.5% (compared to 2.4% in 2025) • A labor force participation rate of 55% (roughly 10 million fewer jobs) • 80% of wealth held by the top 10% (highest since 1939) 🧵 Here's what we found:

20 192 647 380K 709

View Details

Martin Signoux @MartinSignoux

3 months ago

@DFintelligence Exact

0 0 1 113 0

View Details

Martin Signoux @MartinSignoux

3 months ago

@bfmbusiness « Niveau rigueur méthodologique, c’était zéro sur zéro ». (Ne même pas citer le modèle testé alors que c’est ce qui determine TOUT)

1 0 8 110 0

View Details

Martin Signoux @MartinSignoux

3 months ago

This is a really cool project. Worth exploring to get raw impressions from AI usage across the entire spectrum.

Anthropic @AnthropicAI

3 months ago

We invited Claude users to share how they use AI, what they dream it could make possible, and what they fear it might do. Nearly 81,000 people responded in one week—the largest qualitative study of its kind. Read more: anthropic.com/features/81k-i…

807 1K 7K 3.5M 5K

0 0 1 435 0

View Details

Stratechery @stratechery

3 months ago

Agents Over Bubbles Agents are fundamentally changing the shape of demand for compute, both in terms of how they work and in terms of who will use them. They're so compelling that I no longer believe we're in a bubble. stratechery.com/2026/agents-ov…

20 64 460 142K 442

View Details

Martin Signoux @MartinSignoux

3 months ago

👀

Josh Kale @JoshKale

3 months ago

Scientists just copied a Fruit Fly's biological brain and trapped it inside of a computer. Not an AI model trained to act like a fly... A total digital copy of a fly !! This is some sick sci-fi stuff: - They scanned and copied the brain, neuron by neuron, synapse by synapse,

744 3K 19K 2.7M 10K

0 0 0 345 0

View Details

François Chollet @fchollet

3 months ago

When thinking about the societal implications of technology-induced abundance, you can use the present as a reference. The present is already a state of extreme abundance compared to 200 years ago and you should expect similar dynamics to play out

59 54 636 37K 105

View Details

Martin Signoux @MartinSignoux

3 months ago

This is only immersive vibe-coding, but imagine in a couple of years when you’ll be able to create your environment in real-time. You won’t be in a Matrix, you’ll create-as-you-go your own Matrix.

Mistral AI for Developers @MistralDevs

3 months ago

CHECK THIS OUT. Coding in augmented reality with @mistralvibe. Built at the Mistral AI Worldwide Hackathon in Paris by @Min0laa, @tr_mz_, and @glwesteneng. Shoutout to Voxtral and Devstral 2 models. Demo was run on a Meta Quest 3.