Ian

If there is any such thing as "Safe AI", "Dangerous AI" is most assuredly closed source, centralized, and censored.

The new Zenith model just dropped on LM Arena, and early SVG test results are blowing minds.🔥

Hunyuan3D World Model 1.0 is now open-source! A game-changer for game dev, VR, simulation, and digital storytelling.

Synchronicity sent this message to you.

This feels like magic. If we can already control the mouse cursor with our thoughts today, the idea that we will be able to download knowledge into our brains in 10 years' time really doesn't feel particularly absurd.

You should use LLMs locally On your hardware that you own and control Matter a fact, self-host as many things as you can

This Eric Schmidt podcast has a lot of great predictions. must watch. - Super-intelligence will arrive inside 10 years. - AI models will dominate every field within 5 years. - AI models which are World-class AI mathematicians and coders will emerge in 1-2 years, erasing junior…

according to the last LLM you used, you are guilty

x.com/pixhell/status…

Another great @Google paper. 💊 AI crushed the usual clinic bottlenecks here because the team wrapped the model in tight medical guardrails, let real physicians watch everything it said, and then proved the setup works with a head‑to‑head simulation against human residents. 🩺…

Wild. Single shot text to 3D scene (gaussian splat) with no editing. This is approaching a usable scene asset for a game / 3D application.

Accelerate.🚀

Marc Andreessen says we’re living in the barbell era of attention. "Kids want to watch either two-minute videos or three-hour Rogan episodes." TV formats cut every conversation short. Long-form gives people what they’ve been starving for: depth without interruption.

Super interesting talk by Eric Schmidt on AI scaling and AI becoming better than senior engineers and pro-mathematicians in the next 1-2 years. 🧠 Junior coders are already being automated away, and once AI mature even senior engineers who supervise them will no longer be…

Training Jr Artist Jack Fishkin (no public social media yet) that @verenapuhm brought on. Here is one of his experiments today. One lesson was that blurring the mask for the helicopter gave the model nothing to grab onto for the helicopter.

Demis Hassabis says I'm not worried about running out of high-quality human data We already have enough to build strong systems and simulations that generate useful synthetic data "we're already at that stage now"

"Generative Ai will never be useful for professional work", "The topology sucks", "Look at the hands".. You should get used to the idea that things will change in a big way. Embrace it, be creative with the new tools and continue to create beautiful, exciting and surprising art…

It's Friday Night

We're excited to have @shengjia_zhao at the helm as Chief Scientist of Meta Superintelligence Labs. Big things are coming! 🚀 See Mark's post: threads.com/@zuck/post/DMi…

Demis Hassabis says getting AI right could help us solve energy, and lift the limits on human progress. Abundant energy makes water trivial, infrastructure cheap, and expansion to the stars inevitable. “Bringing consciousness to the universe. Waking up the universe.”

Music is math.