
Co-founder @BovoMusic • @Salesforce-acquired 2X Founder • Ex-@NVIDIA • Co-author @ManningBooks Designing DL Sys
United States
Joined September 2010
- Tweets1.2K
- Following658
- Followers688
- Likes668
Jensen's saga is a Chinese AI drama IRL:
- Top of the world
- Scandal
- Ousted by the board
- People wants him back
- Board and gov say no
- Market crash
- "We've got it" "Let's move forward"
1
54
Having fun with the Bovo Violin and the 2nd movement of Butterfly Lovers! Butterfly Lovers has a special place in my heart because it was my first public performance in front of a thousand-strong crowd.
Get the plugin at bovo.ai/instruments/violin?u…
@BovoMusic
2
71
Donald Szeto retweeted
Bovo Violin 0.2.20 is out! 🎻
You asked, we listened.
Based on user feedback, we've expanded compatibility to Windows 10 and macOS 12 Monterey, with testing across more DAW hosts.
Free physically modeled violin. No sample library.
Download: bovo.ai/instruments/violin?u…
I wonder how much "increase" can make up the value that's already erased from these employee's portfolio. Sounds like a death spiral to me.
Interim CEO Collette Kress sought to ease employee concerns following Jensen Huang’s ouster, in part by promising to increase compensation across the board. buff.ly/tv6OjUQ
1
43
This is the best demo of a person who's extremely good at developer experience and product. It's a rare combo.
Rex demo time: magical and automatic remote copy/paste, drag/drop, and port forwarding (e.g. open web servers running on remote machines in your local browser). Plus, showcasing some of the internal plumbing that powers this and how you can hook into it with Lua scripting.
Our goal with Superlogical and Rex is for everything to feel local. I've said before: the moment any UI shows a selector between "local" and "cloud", the product has failed. The lines should blur, and we should build the foundational technology that makes it happen performantly, securely, and beautifully.
1
78
We built a physically modeled violin with GPT-6.2 Astra…
Well, its predecessor. And a lot of physics, modeling, and calibration.
Bovo Violin synthesizes sound in real time, with continuous control over bowing, fingering, and vibrato. No sample library.
Here's the full performance of Bach's Partita No. 3 (Preludio), not just a short excerpt.
Free download: bovo.ai/instruments/violin?u…
Feedback welcome from musicians and audio developers.
A physically modeled violin playing Bach's Partita No. 3, BWV 1006. The complete performance, not just a short excerpt.
This is Bovo Violin, our new free virtual instrument. No sample playback—the sound is synthesized in real time, with continuous control over bowing, fingering and vibrato.
Free download:
bovo.ai/instruments/violin?u…
1
148
This is accurate. I was at NVIDIA when CUDA and G80 just came out. When I left the company because my naive self had doubts in its mobile strategy, I still thought sustaining CUDA was Jensen's best decision ever.
Replying to @davextesla @JeffreyC1409
I was in NVIDIA’s core GPU design group and can attest to how much Jensen’s strategy shaped the company today.
We laid the foundation for compute that formed the basis of the industry before it even existed.
AMD also tried to introduce GPGPU via OpenCL instead of CUDA. OpenCL, formed by committee, had a slow iteration cycle, while CUDA addressed real customer problems. Pressing needs were quickly met by the new architecture - better cache, instructions, ALUs, or warp granularity.
CUDA did not take off for years. AMD saw little value in GPU computing, abandoned it quickly, and focused on extracting revenue from declining gaming markets.
Investors questioned why NVIDIA did not do the same by producing more lower-end graphics cards instead of pursuing parallel computing, where the CPU dominated.
Without Jensen, CUDA would have met the same fate, and people would still rely on suboptimal committee-developed tools.
2
65
I have new found respect for Mark Zuckerberg… for now
Hoping my friends at NVIDIA will survive this tide
1
61
Be meticulous reviewing your agent's plan. 15 minutes of reading can save you 15 hours down the road.
39
Thursday, 3:50pm, Silicon Valley. 412 orders ahead. 206 minute wait. For a kale avocado smoothie. AI agents are writing the code now, so engineers finally have time to stand in line for green juice. Real productivity gains.
92
Congrats! Looking forward to a Codex that doesn't break almost every week!
Some news: I’m joining OpenAI to lead product management for Core Products!
I have a lot of respect for OpenAI for showing the world that scaling works, for bringing AI to over a billion people, and for believing great products are core to the mission.
I’m excited to work with @thsottiaux, an incredible PM team, and everyone across OpenAI. Together, we’ll build agents that have your back and your team’s, and keep you in control, so you get more time for the people and work you love.
Lots to learn, and lots to build. Please keep your feedback coming!
1
59
Tried the popular "roast me based on everything you know about me" prompt, and this one hit a little too close to home. 😂
Despite multiple appeals and even speaking directly with @Reddit staff, my account is still banned to this day.
47
"You're invited to try out Rex from Superlogical"
This just made my day. Thanks @superlogical @mitchellh !
36
In the AI era, you can pretty much blame everything on rogue AI agents
Hey @Railway, is there a reason employee's of yours are trying to breach my infrastructure?
Your employee Jake Runzer, just attempted to tonight.
35
Bro just found out the secret sauce of frontier labs. Frontier models can reprogram LLM's network directly.
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
2
82













