Inlay

How big a leap is Mythos in cyber capabilities? Timothée Chauvin, Alexander Barry, Jean-Stanislas Denain, and Anson Ho compiled the public evidence and found that while it’s unclear if Mythos was ahead of trend in discovering vulnerabilities, it represents a big jump in exploiting them. 🧵

Claude Fable 5 scores very well on FrontierMath: Tiers 1–4 (v2), reaching 87% on Tiers 1–3 and 88% on Tier 4. This continues a streak of Anthropic models improving rapidly at math.

I think it is really worth reading this piece on RSI at Anthropic. There is a bit of navel-gazing, some marketing, and a lot of very sincere beliefs about what Anthropic thinks is likely in the near future of AI that you probably want to be aware of. www.anthropic.com/institute/re...

If leading AI companies are indeed approaching the point of recursive self-improvement, a coordinated, verifiable, and universally applied pause is probably the only responsible solution to mitigate several major AI risks; at least until safety guarantees are developed and demonstrated. (1/2)

Arrêtez la bidoche. C'est une catastrophe aussi bien éthique qu'écologique. On peut se nourrir autrement. Ce n'est ni plus difficile ni plus coûteux (contrairement à ce qu'on entend souvent). "Décarner" son alimentation est l'acte individuel le plus impactant sur le plan écologique (et politique).

1/ Étienne Klein a réagi à l'annonce de son retrait de doctorat en publiant un texte quasi similaire à celui publié - puis effacé - en août 2025 (un an après la publication de notre enquête initiale). Je repartage ici - en l'adaptant - un fil de réponses que j'avais fait à l'époque.

Surprise surprise! It is WHAT is processed and not HOW or HOW MUCH it is processed that matters for health. This is EXACTLY the argument that Gabriel Rosenberg and I made in Feed the People! using many of these same studies. www.science.org/doi/10.1126/...

I’ve been writing a lot of long deep dives recently and wanted to step back and write a quick rant about why I’m consistently landing so hard in the skeptic corner in the data center debate blog.andymasley.com/p/why-i-thin...

Mon test usuel à la sortie d'un nouveau modèle c'est l'écriture de nanofictions. Avec seulement quelques exemples dans le prompt (et pas de sujet imposé), voici les trois premières que Claude Fable ait produit:

Announcement that ‘policymakers’ need to be convened by US firm viewed as marketing ploy by some experts Anthropic has floated the idea of a worldwide “temporary pause” on AI development – and said it was going to convene “policymakers” to discuss the dangers of advanced AI – in its latest release touting the capabilities of its products. In a long post on Thursday, Anthropic detailed the progress of its AI model, Claude, towards “recursive self improvement” – that is, being able to make better and more powerful versions of itself. Recursive self-improvement is a bugbear of AI safety researchers, viewed as the key step for AI to become superintelligent and therefore unleash widespread consequences on humanity. Continue reading...

Premier commentaire public de l'intéressé, version remaniée et enrichie d'un texte posté (puis supprimé) à l'été 2025. L'article de @arretsurimages.net coécrit avec @jeanabbiateci.bsky.social a été mis à jour en conséquence : www.arretsurimages.net/articles/the...

(1/4) Nombreuses sont les personnes qui aiment réfléchir avant de juger. Ces quatre petites pages sont pour elles, exclusivement pour elles. Elles doivent être lues jusqu’au bout. Bonne journée à tout le monde.