AI is literally tempering with our timeline.

x.com/i/article/2029…

You are playing 7 new ball games the moment you move your AI models from sim to real. Real Physical Robots. And they matter. unless you teleoperatinging with a controller then well...

Researchers tested AI against 100,000 humans on creativity. nature.com/articles/d4158… AI beat the average person. That's the headline everyone will run.

This is why we're building Personal Intelligence at Onix. The future isn't AI replacing experts. It's AI making the best experts unreachable. The real question isn't "will AI replace humans?" It's "which humans will use AI to pull even further ahead?"

x.com/i/article/2029…

A robot that crushes a benchmark in simulation or on a pristine test track can (and often does) fall apart the moment it leaves the lab. SiHing Guppy technical article here for those who are interested in understanding how perturbations introduced in robotic system can impact

Nobody pays for "AI knowledge" in 2026. They pay for one clear outcome. The creator economy figured this out before Big Tech did. The experts winning right now aren't the ones with the biggest following. They're the ones with the deepest skill in the narrowest lane.

furthest thing youtu.be/9j6VAg_QlvQ?si…

x.com/i/article/2031…

SiHing Guppy leading a team of ML engineers doing Perturbation Evaluation to BREAK Assumptions in Robotics.

you cross the t’s and i dot the i’s

Many talented people over the past few years were declined an offer or even an interview @xAI. My apologies. @BarisAkis and I are going through the company interview history and reaching back out to promising candidates.

We’ll be at NVIDIA GTC next week in San Francisco. SiHing Guppy and Marc Alloul will be there talking about model transparency and evaluation for robotic policies: how to understand what your models are actually doing before you deploy them. If you’re building or deploying

Nvidia GTC find our team there and we are happy to share how we pushing innovation in AI / Robotics field.

Yes

constant skillful cog abstra

x.com/i/article/2033…

Language removal: –1 % success. Vision removal: –82 %. These models are vision-dominant shortcuts, not true language-grounded agents. Does this mean today’s benchmarks (and the valuations built on them) are mostly crediting the wrong thing?

Lil Wayne speaks on AI in music: “My friends were a little worried, they said ‘You can ask [AI] to do a verse like Lil Wayne’ and so I did it: ‘Let me hear a verse like Lil Wayne’…I did it on a couple devices…They all sucked. We gon be okay”