ᴊᴀʏ ᴄᴀʀʟsoɴ retweeted
Last week, NVIDIA spun out DAQIRI, the library we use at the Allen Telescope Array to process 200 Gbps of UDP packets per GPU/ConnectX pair in real time.
I'm planning to release a CyberEther Plugin for it soon!
Check it out: developer.nvidia.com/blog/en…
Amazing new chapter for our friends over at MJ. Literally turning science fiction into reality. All of us at NV are excited to be along for the ride!
Vienna-bound for ICRA 2026! Excited to talk world foundation models → real-world deployment and how the runtime architecture has to evolve with them. If you're there, let's meet up 🤝 nvidia-holoscan.github.io #ICRA2026 #Holoscan
Hey! Our team is hiring. We're looking for a jack-of-all-trades DevRel who's interested in sensors, robotics, stupid-fast Ethernet, and helping our partners build amazing, safe products needed for the next generation of robotics.
nvidia.wd5.myworkdayjobs.com…
Couldn't be more excited for #CES2026! I'll be talking to robotics, semiconductor, and sensor manufacturers about how NVIDIA Holoscan and Holoscan Sensor Bridge are unlocking the next generation of software-defined sensing across medical, auto, industrial, and scientific domains.
Get ready, Las Vegas.
NVIDIA Live with CEO Jensen Huang at @CES is about to set the stage for what’s to come.
Don’t miss your chance to see how the next generation begins—watch it live.
📍Fontainebleau, 2nd Floor. Doors Open at 11:30 AM PT.
(No CES badge required).
🎥Livestream available
RSVP here:nvda.ws/4pl9pRv
#CES2026
ᴊᴀʏ ᴄᴀʀʟsoɴ retweeted
New light functionality for if you want to pick up the opponent's piece to be taken first, a few bugs to resolve with it but it's a good start
ᴊᴀʏ ᴄᴀʀʟsoɴ retweeted
You know something is at the limit of physics when pin assignments alone will cause 1GHz bandwidth loss.
ᴊᴀʏ ᴄᴀʀʟsoɴ retweeted
Any recommendations on debugging UDP timing or really anything in freertos? Breakpoints seem like a bad play, so between println debugging, toggling gpio pins and a logic analyzer, and Wireshark, I've got a lot of info, but it's been hard to line up when different things occur
This is definitely my vice president
I'm obsessed with @Tim_Walz singing the Menards theme song and then ranting about keeping your receipts to get the rebate. Incredible content.
I love working at a company where it's completely normal for entry-level embedded software engineers like @carlolson_dev to have to bodge 800 Mbps MIPI DSI traces that us EEs screwed up
This is a great technique for high-density designs that's quite common on cell phone PCBAs
"Or, while it pains us to give you this option, we're legally required to let you opt-out of all these email lists that we decided to automatically sign you up for the moment you registered an account on our site"
This is going to be a huge deal in so many industries
ftc.gov/news-events/news/pre…
Next time I accidentally swap SDA/SCL for the zillionth time, I'll just tell myself I'm testing the IC for alternate address modes
While the C abstract machine can map somewhat cleanly onto something like a Cortex-M0 microcontroller, writing performant C code for a modern out-of-order-exec multilevel-cache SIMD-capable desktop CPU is about as "close to the metal" as programming in Java
Most C code gets shoveled into a compiler with -O2 ("because Steve said -O3 made the program crash on Tuesdays"), and no one thinks anything of it, but there are like 2734981327592 transformations your compiler is secretly doing. Your compiler is the Java VM.
To become a Real C Developer™ you spend 100x more time reading the compiler manual than you spent learning the language, so you know all the #pragmas and __attribute__(())s you have to plaster everywhere to make your code work, and if anyone ever suggests switching to Clang...





















