Great launch AI Foundations Engineering Manager @Ryfeus
’ take: "At Snorkel AI, we are focused on building frontier datasets. Our work spans the full AI lifecycle, from curated datasets like Snorkel Data Series for the most challenging agent tasks to evaluation, reinforcement
RESEARCH
-
Snorkel AI Launches Frontier Dataset Engineering Manager Role
By
–
-
Mythos Could Transform Security Research Like Hubble Did Astronomy
By
–
Mythos could do for security research what Hubble telescope did for astronomy.
-
Claude Mythos Preview: Powerful AI Model Analysis and Safety Concerns
By
–
¡NUEVO VIDEO en el LAB! Claude Mythos Preview nos ha pillado por sorpresa aún cuando estamos acostumbrados al ritmo de progreso de la IA. Un modelo tan potente como peligroso hasta el punto de que no verá la luz… Hoy analizamos esta nueva bestia! Link a continuación
-
LLMs Cannot Learn Temporal Reasoning Due Architectural Limitations
By
–
To anyone who read Rebooting AI back in (checks notes) 2019, this is both hilarious and unsuprising. The field has wasted 7 years on an architecture that can’t solve one of the most basic litmus tests Davis and I discussed then: temporal reasoning. Lavanya (@naynaval) LLMs cannot be taught temporal context due to the way they are architected. How do you teach temporal understanding to probabilistic models hosted across distributed data centers?? 🤷🏻♀️ — https://nitter.net/naynaval/status/2041754297491398794#m
-
Safetensors joins PyTorch Foundation for AI model security
By
–
Glad to see that Mytos is putting back cybersecurity in the spotlight. Our main contribution to the topic has been the creation of Safetensors by @narsilou @huggingface in collaboration with @AiEleuther @StabilityAI.
— clem 🤗 (@ClementDelangue) 8 avril 2026
It has become the most popular way to share models safely and… pic.twitter.com/ind7o6R6i6Glad to see that Mytos is putting back cybersecurity in the spotlight. Our main contribution to the topic has been the creation of Safetensors by @narsilou @huggingface in collaboration with @AiEleuther @StabilityAI. It has become the most popular way to share models safely and today @LysandreJik announced at #PyTorchCon that the project is joining the @PyTorch foundation to continue to scale (for example with torch core integration). Let's turn talks about cyber-security into action!
-
Humachineology: Evolving Technohuman Competencies Framework
By
–
Humachineology and humachinekind provide a framework to study how technohuman competencies have been evolving over time.
-
Anthropic Leak: Building Resilient Systems Through Open-Source Security
By
–
Anthropic had the most powerful cyber-security model in the history of this world and their internal code based still leaked? We should assume everyone can be compromised, and build systems that keep the cost of attacking higher than the reward, limit blast radius when attacks succeed, create fast repair loops after weaknesses are found and reduce systemic risk. Open-source will play a major role in all of that!
-
Real-World Multi-Step Workflows Beyond Isolated Agent Benchmarks
By
–
The agentic benchmarks are interesting but I want to see how it performs on real multi-step workflows not just isolated tasks
-
Major News Sites Undercover Claude Mythos Release Coverage
By
–
Think about it. None of the major newspapers have given Claude Mythos a prominent mention/Top Placement on their websites. We live in an AI ivory tower; 99% of people don't know what was published yesterday. They don't know what's coming. That's why our work is so important. Shakeel (@ShakeelHashim) The Anthropic Mythos release does not appear near the top of the homepage on any major news site today. The NYT is closest, but it's still pretty far down. The Guardian thinks a Vogue cover with Anna Wintour and Meryl Streep is more important. The Washington Post is prioritizing yet another "we tried to get into Berghain" story. The media is not adequately covering the insane moment we are in. — https://nitter.net/ShakeelHashim/status/2041829164894871584#m
→ View original post on X — @kimmonismus, 2026-04-08 14:04 UTC
-
GLM-5.1 Real-World Usability Compared to GLM-5
By
–
There is no reason to believe that they made GLM-5.1 worse than GLM-5 in real-world usability?