Amazing to see @nvidia go all-in on Codex! We’ve started going deep with a few companies to roll out Codex and GPT-5.5. Seeing it work this quickly at this scale is exciting. Let’s fire up those Blackwells!
AI HARDWARE
-
Mac Pro Struggling Under AI Agents Computational Demand
By
–
I never particularly cared about the performance of my laptop, but now I'm running my mac pro into the ground with all of these agents. We need so much more powerful compute, this is insane.
-
Intel Stock Investment Decision in AI Era
By
–
I was just thinking a couple of days ago – should I buy Intel stock? Nah, I thought, surely it's all priced in by now
-
Compute Infrastructure Costs Vastly Exceed AI Model Training Expenses
By
–
perhaps interesting to think through that compute infra building is probably a 0.5-1T problem, but training, if cursor is all it takes, is just a 0.06B problem
-
Synthetic Aperture Radar: Satellite Tech for Global Monitoring
By
–
The same satellite tech can track a convoy through dense cloud cover at night, flag a sinkhole weeks before it opens, catch a village built on a collapsing slope, and even map the inside of a pyramid from orbit.
— Bilawal Sidhu (@bilawalsidhu) 23 avril 2026
It's called Synthetic Aperture Radar (SAR), and a lot of the data… pic.twitter.com/vz4xBT18MJThe same satellite tech can track a convoy through dense cloud cover at night, flag a sinkhole weeks before it opens, catch a village built on a collapsing slope, and even map the inside of a pyramid from orbit. It's called Synthetic Aperture Radar (SAR), and a lot of the data
-
27B Model Running Locally on 5090 Beats Frontier Model
By
–
A 27B running locally on a 5090 beating a frontier model is where 2026 actually gets interesting.
-
NVIDIA OpenAI Partnership Reduces Token Costs by 35x
By
–
Efficiency isn't just about speed anymore — it's about the massive reduction in the cost of intelligence. NVIDIA and @OpenAI
's partnership leverages the GB200 NVL72 to deliver a 35x reduction in token costs, bringing enterprise-grade AI to an unprecedented scale. Trained and -
BrainCo Revo 3: Revolutionary Robotic Hand with 21 Degrees of Freedom
By
–
The most human robotic hand ever built just dropped.
— AI Highlight (@AIHighlight) 23 avril 2026
BrainCo Revo 3 has 21 active degrees of freedom, 70N grip force, and pressure sensitivity down to 0.01N.
Direct-drive architecture gives every joint independent precision control with zero mechanical stiffness.
Full-palm… pic.twitter.com/jyHBAR50sgThe most human robotic hand ever built just dropped. BrainCo Revo 3 has 21 active degrees of freedom, 70N grip force, and pressure sensitivity down to 0.01N. Direct-drive architecture gives every joint independent precision control with zero mechanical stiffness. Full-palm
-
Google TPU 8i Co-Designed for Low Latency Inference
By
–
TPU 8i is co-designed with our Gemini research team to support low latency inference. Among the attributes that support this are large amounts of on-chip SRAM, enabling more computations to be done on chip without having to go to HBM for weights or KVCache state as often. The
-
TPU 8t: 3X FP4 Performance Boost Over Ironwood
By
–
First, let's talk about TPU 8t, which is designed for large-scale training and inference throughput. The pod size is increased slightly to 9600 chips, and provides ~3X the FP4 performance per pod vs. Ironwood (8t has 121 exaflops/pod vs. 42.5 exaflops/pod for Ironwood). In