NEW DEEPSEEK MODEL RELEASE 🚨 Two models: v3.2 - GPT-5 level performance v3.2 speciale - Gemini-3.0-Pro level performance, gold medal level for IMO, CMO, ICPC World Finals & IOI 2025 GRPO with modifications: - unbiased KL estimate - off-policy sequence masking - keep MoE

Got burned by an Apple ICLR paper — it was withdrawn after my Public Comment. So here’s what happened. Earlier this month, a colleague shared an Apple paper on arXiv with me — it was also under review for ICLR 2026. The benchmark they proposed was perfectly aligned with a

Jensen right now...

1/ At the core is a new American Science and Security Platform. DOE is ordered to turn the national lab system into an integrated stack that provides: • HPC for large-scale model training, simulation, inference • Domain foundation models across physics, materials, bio, energy

iykyk

Today, our team launched Google Antigravity. - Agent-first IDE powered by Gemini 3 Pro 🧠 - Browser control to test your apps automatically 🤖 - Agent Manager to orchestrate parallel agents ♾️ Stoked to keep shipping with the @antigravity team. This is going to be fun.

After a long time I found myself shopping for a tech device that I am unfamiliar with. Instead of browsing reddit, I hopped on various LLMs for “research”. Gemini is so incredibly delightful that I left a positive feedback to that chat session. I almost never gave positive

🤔🤔

It took one year to build me

i got an idea

I found @cursor_ai's composer model extremely useful for many specific tasks. A very short feedback loop means I can prompt it 4-5 times, and it can instantly get to the code I wanted. Speed is really a 10x factor in AI

Jensen Huang: “Are there any NVIDIA, $NVDA, investors here?” (Audience cheers) Jensen Huang: “That’s why Korea is rich.”

Grok. Best response so far

Sonnet 4.5 got the right idea and started streaming a valid response. But, most likely due to copyright restriction filters, decided to abruptly stop generating

Haiku 4.5 fails abysmally even after prompting

“Thinking” killed the fun. In the past ChaGPT was playful enough to catch on to lyrics. Now… (Song: Friday, Rebecca Black)

RIP fine-tuning ☠️ This new Stanford paper just killed it. It’s called 'Agentic Context Engineering (ACE)' and it proves you can make models smarter without touching a single weight. Instead of retraining, ACE evolves the context itself. The model writes, reflects, and edits

Breaking the Code: Security Assessment of AI Code Agents Through Systematic Jailbreaking Attacks Code-capable large language model (LLM) agents are increasingly embedded into software engineering workflows where they can read, write, and execute code, raising the stakes of

Meta just ran one of the largest synthetic-data studies (over 1000 LLMs, more than 100k GPU hours). Result: mixing synthetic and natural data only helps once you cross the right scale and ratio (~30%). Small models learn nothing; larger ones suddenly gain a sharp threshold

Hamas agreed to release all Israeli hostages captured in its Oct. 7 attack and accepted other parts of US President Donald Trump’s plan to conclude the war in Gaza. Listen to that and the latest stories on Bloomberg News Now.