another @raais alum returns! excited to host @vivnat – he leads medical and scientific ai at @GoogleDeepMind – from Med-PaLM to AMIE to the AI co-scientist…some of my favorite work his last talk was in 2023, the early days of LLMs in healthcare come along for the latest!
LLMS
-
Update soul files specifically for GPT-5.4
By
–
Getting much better. Make sure to update the soul files specifically for Gpt-5.4
-

LLMs Assessment for Clinical Diagnosis: Safety and Reasoning Limitations
By
–
Assessment of 21 LLMs for generating a differential diagnosis
"Off-the-shelf LLMs have not yet achieved the intelligence required for safe deployment and
remain limited in demonstrating advanced clinical reasoning." https://
jamanetwork.com/journals/jaman
etworkopen/fullarticle/2847679?utm_campaign=articlePDF&utm_medium=articlePDFlink&utm_source=articlePDF&utm_content=jamanetworkopen.2026.4003
… -

Harvard Study: AI Safety Models Withhold Critical Medical Advice
By
–
Harvard just proved the "safest" AI models cause the most medical harm. AI safety models refuse life-saving medical advice to patients. Then give it freely when you pretend to be a doctor. A new benchmark tested 60 medical emergencies across six major models. Same clinical
-
Models as Components: Infrastructure and Orchestration Matter Most
By
–
My biggest takeaway: Models are becoming components, not products. What matters now is the system around them: → runtimes
→ memory
→ tool access
→ orchestration
→ secure execution environments That is what turns a model into an agent, and an agent into something the -
Hybrid Strategy: Combining Proprietary and Open Models
By
–
The second misconception I keep seeing: Too many teams think they have to choose between open models and proprietary models. They do not. The smarter path is hybrid. → proprietary models for scale and broad capability
→ open models for flexibility and control
→ -

Claude Mythos Hacking Capabilities and Security Implications
By
–
This is yet another example of Claude Mythos’s incredible hacking capabilities. I expect we’ll see more examples and independent evaluations in the coming weeks that make clear just how powerful (and dangerous, in the wrong hands) this model could be.
-

Mythos Performance Comparison with Opus Models
By
–
So looks like Mythos is better, but not an alien model – jump between Opus 4.5 and Opus 4.6 was similar to a jump from Opus 4.6 to Mythos Preview
-

Claude Mythos Explained on CBS Mornings Without Jargon
By
–
I went on @CBSMornings to break down what Claude Mythos means for all of us.
— Matt Shumer (@mattshumer_) 13 avril 2026
If you want to understand the latest in AI, without the jargon, give it a watch! https://t.co/6OXIRCGr5xI went on @CBSMornings to break down what Claude Mythos means for all of us. If you want to understand the latest in AI, without the jargon, give it a watch!
-

Opus 4.7 and Sonnet 4.8 Releases Imminent
By
–
Quick reminder, that Opus 4.7 and Sonnet 4.8 releases should be imminent as well.
