I’m convinced now 👀
We’ll see a fully autonomous AI researcher by the end of this year. Choosing what to investigate, running experiments and working through failures without a human steering every step. Astra is making that timeline hard to dismiss.
Expertise is no longer only about knowing the right answer.
In AI-driven work, it also depends on asking better questions, reading context, and adapting judgment when conditions change.
By @antgrasso.
Recently, I’ve been thinking about which fundamental skills are worth developing for the AI-native age.
One observation stuck with me: many people can speak fluently; but much rarer is the ability to convey a deep idea in just a handful of precise words.
Congrats @micoolcho@chris_j_paxton@DJiafei and thanks for having me as the #3 guest! Was fun chatting about sim-to-real dexterous manipulation 😄
Amazing to see how much the AI robotics community has grown since then. Look forward to the next 100 episodes!
Greg Brockman is now openly saying GPT-6 Astra may be AGI. 🤯
He literally ended the briefing with: “Welcome to the AGI era.”
And Astra/Aeon access is coming to users in the next few days.
This is getting very real.
DLR and TUM developed eBert, a 5.2 kg, 12-DOF quadruped that uses the physics of its own body to generate locomotion.
Researchers identified six nonlinear motion modes in the robot’s spring-loaded mechanics, then excited those modes with a relatively simple controller.
The result was naturally emerging gait patterns including pacing, trotting, bounding and hopping.
In some modes, the springs supplied up to 80% of the positive mechanical work. Four of the six predicted modes were successfully reproduced on the physical robot.
A strong example of how robot morphology and compliant mechanics can reduce how much locomotion has to come from active control.
Source
https://t.co/m6jGx9wnO5
What actually happens when a neural network learns?
It starts by making a prediction. The difference between that prediction and the desired answer becomes the loss. Backpropagation then works backward through the network, using the chain rule to determine how much each weight and bias contributed to that error. An optimizer uses those gradients to adjust the parameters, and the process repeats, often millions of times.
$NVDA told JPMorgan that “without supply constraints, the business could more than double YoY” even against its already aggressive ~70% FY28 growth estimates.
The more interesting part is why Nvidia gave that outlook so early since management believed Street estimates were low enough that they could actually distort supply-chain planning.
So FY28 guide wasn’t just about resetting investor expectations but was also a message to the semiconductor ecosystem to start planning for a much larger demand curve.
#OpenAI will “soon” launch its most advanced #cybersecurity AI model yet, setting up the release of a direct competitor to Anthropic’s Mythos model.
#Forbes
For More Details: https://t.co/FZh8QNxrlL
Fatigue and tensile behavior of 3D-printed polylactic acid matte: Effects of infill pattern and density. https://t.co/jDGT4lfsP2
#additivemanufacturing
Stay up to date with the latest news from the electronics industry and read expert insights and in-depth technology reports in the latest edition of our #EIU magazine. 📰
This issue focuses on the latest developments of energy-efficient design.
Read now: https://t.co/pxARPSy4en
Here’s the first podcast I’ve ever done! It was a joy to speak with Vatsal and Jake.
We discuss the neuroscience that underpins my research, how grokking works, and what AI means for the future of human creativity. Much of this I haven’t spoken publicly about until now.
What could your team accomplish if specialized workers stayed focused on production instead of moving materials?
One commercial print manufacturer reduced cycle times by 97% by automating material movement. Discover more in our Sept. 16 webinar: https://t.co/QdvAlWvuML
MediaTek and @nvidia today announced an expansion of their collaboration across AI infrastructure, local AI computing and automotive.
MediaTek will adopt the NVIDIA NVLink Fusion platform, providing hyperscalers, cloud service providers and frontier model developers a more scalable, flexible and faster path from custom XPU design to rack-scale AI factories.
August, but make it productive: Automate drawing creation from MBD data, create rigid-flex PCBs, streamline sheet metal design with the new Jog feature, and much more. Catch up on the latest Onshape release: https://t.co/25j0OwUIx1
Autodesk 2026 AI Jobs Report: AI hiring in Design and Make more than doubles as students face a new skills gap https://t.co/E7hwbSMPpe
#AutodeskExpertElite