Several people encouraged me to share a post I made on LinkedIn, here on X
I wrote this from the heart, about someone I had the privilege of working closely with for years.
No agenda. Just a side of Jensen I hope doesn’t get lost in the noise.
As the days pass, my heart grows heavier.
I have worked at NVIDIA for nearly eight years. I have never worked for anyone quite like @JensenHuang , and I doubt I ever will.
Much has been written about his extraordinary vision and the company he built. But there is another side to Jensen that I have had the privilege of witnessing. And I am only speaking to what I have personally witnessed.
I have watched him stand behind researchers, founders, and startups when few others would. Companies struggling to survive, ideas that seemed too early, people whose ambitions exceeded their resources. He saw possibilities most of us could not yet comprehend. And he committed his time, attention, and NVIDIA’s resources to helping them succeed. Some of these companies might not exist today without him.
His capacity for long-term thinking is extraordinary. Sometimes it took me years to appreciate the implications of something he had recognized almost immediately. Few could match his ability to connect technology, markets, and people.
My own journey at NVIDIA has not always been easy. Through difficult organizational transitions and moments of uncertainty, Jensen’s guidance and willingness to stand behind my work helped me persevere.
But what I will remember most is his humanity.
My sister, has been fighting an unforgiving battle with brain cancer. When her condition deteriorated during her visit to California, our family faced unimaginable circumstances: devastating medical decisions, overwhelming hospital expenses, and the possibility of losing her far from home.
Despite everything on his shoulders, Jensen found the time to care. He helped my family through circumstances that seemed impossible. And he took the time to meet my sister.
The photograph here is one I will treasure for the rest of my life.
When he told me, “I am proud of you. Treasure this time together,” I understood something about leadership that no professional experience could have taught me.
What I will miss most professionally is the extraordinary combination of vision, conviction, curiosity, and personal commitment he brought to everything. His ability to see around corners, to challenge conventional thinking, and to bring an entire ecosystem along with him.
Organizations evolve, leadership changes, and new chapters begin. But the influence of someone who has shaped so many people, ideas, and possibilities endures far beyond any role or title.
Jensen built far more than NVIDIA. He helped build an ecosystem of researchers, entrepreneurs, engineers, and companies whose futures became intertwined with his vision.
I do not know what comes next. I only know how grateful I am to have worked with him, learned from him, and experienced his kindness when my family needed it most.
Thank you, Jensen. For your guidance, for believing in people and ideas long before the rest of us understood them, and above all, for your extraordinary humanity.
Perplexity at @sgl_project summit: CAKE 🍰 generates the best CUDA kernels at mass, accelerating training & inference at industrial scale, by @abcdabcd987
So glad to have the opportunity to share our team's recent work of rebuilding a training framework at #SGLangSummit. We'll share blog posts and open source the project very soon. Stay tuned 😜
So glad to have the opportunity to share our team's recent work of rebuilding a training framework at #SGLangSummit. We'll share blog posts and open source the project very soon. Stay tuned 😜
Databricks just took 2nd place in Meta's Muse Spark Open Kernel Challenge, out of 300K+ submissions!
Our kernels achieved an average 2.5X speedup over SGLang kernels.
We built a Kernel Factory where AI agents autonomously hill-climb their way to better kernels. We found Opus 5.5 to be the strongest model for GPU kernel optimization.
These kernels go straight from our Kernel Factory into production, powering Databricks Unity Gateway and AI inference for our customers.
1/n
Introducing our NeurIPS’26 work 💙Evolving Agent Teams (EAT) 💛, where a supervisor diagnoses why an agent team stalls and restructures it to escape local optima, for long-running optimization tasks.
EAT shows superior performance across domains:
🏆 Open Kernel Challenge, decode layer on H200: 2.449× over SGLang, #4 of 1,180 (as of Nov 1, 2026)
⚡️ NVIDIA kernels: KDA prefill on B200, 3.44× over FlashKDA (previous best: 2.96×)
🔥 AMD kernels: KDA prefill on MI350X, 3.01× over vLLM
💻 MoE serving on Mac: Qwen3.8-Flash-Next (Unsloth UD-IQ4_XS GGUF) on M5 Pro, ~3.3× faster prefill (930 tok/s) and 1.22× faster decode (43.7 tok/s) than Strata at the same memory
🦾 RoboDojo: 62.1 on held-out seeds (the best leaderboard entry as of Oct 26, 2026 is 55.3)
- Paper: https://t.co/KeSz8RyTne
- Code: https://t.co/cdLATp2jpw
🧵👇
New on our Frontier Red Team blog: using Claude to formally verify software at large scale.
Formal verification can prove that code is free of certain bugs, but historically it has only been feasible for small codebases (or required a lot of expert time).
With AI models, however, this is beginning to change. In collaboration with @theoremlabs, we used Claude to prove security properties in over one hundred real-world open-source libraries over the span of a few weeks.
Read our research in full here: https://t.co/zUyx8CjBEk
So happy to see this 🎉 Glad to have contributed the specialized decode path (vllm#61868, #61869) behind
these results. On the same GB300 it brings +36% P90 interactivity for a single user. Thanks @vllm_project, and more to come!
https://t.co/wPHZGBFOR1
https://t.co/PG7nQZLr8d
The AI industry has a massive compute efficiency problem. At Physion Labs, we think it's time to challenge it.
Today, we're sharing the technology behind NexTron, our world-intelligence training engine.
Our benchmarks on matched hardware:
⚡ Up to 20.9× SFT throughput vs. DeepSpeed (H100)
⚡ Up to 6.6× SFT throughput vs. Megatron (H200)
⚡ Up to 7.2× SFT throughput vs. Megatron (B200)
⚡ Up to 3.1× end-to-end RL throughput vs. Miles–Megatron
🔥 180B model fine-tuning on just 8 H100 GPUs.
Same hardware. Dramatically higher throughput. Comparable model quality.
We believe the next frontier isn't just building bigger GPU clusters.
It's unlocking more intelligence from every GPU, every watt, and every dollar spent on training.
#PhysionLabs #NexTron #WorldIntelligence #AIInfrastructure