I got my first Twitter account in June 2006. I was user #314. I got this account in 2007. This is my new main account.

Interestingly, this is part of why people thought AI training would fail as it began to ingest its own data - "model collapse" - but it turns out clever engineering has helped avoid this problem and, indeed, AIs are now mostly trained on AI-generated "synthetic" data.

21/ An AI Content Audit Assistant analyzes sites with MCP tools. @chris_nectiv combines Screaming Frog and Zapier.

A multirobot system, developed by a Caltech research group and TII in Abu Dhabi, has a Unitree G1 humanoid launch a transforming drone from its back.

🇨🇳 China just rolled out a mass-produced grenade drone. Foldable props. Infantry-ready. The battlefield just changed. Unlike most repurposed quadcopters, this model is designed from the ground up for combat. It can carry grenades, fold into a backpack, and integrate with standard infantry gear. The design focuses on: ✅ High-lift propellers for heavy payloads ✅ Foldable arms for portability ✅ Swarm compatibility and remote targeting support What makes it notable is not just the specs. It is scale. This is not a prototype or a one-off experiment. It is part of a mass-production initiative for light combat drones. Low-cost. Modular. Disposable. That is where small-unit warfare is heading. A glimpse into how drones will reshape tactics at the squad level. Fast, lethal, and autonomous. Saw this at @Defence_Index

This is not perfect, but it also doesn't seem like a model trained on video should be able to get so many details of the dynamics right: "honey pours down a marble statue of a toucan, the nose cracks and falls off"

AI video models may not be complete world models, but they are oddly capable of fairly sophisticated (if flawed) "simulations" of novel situations. Veo 3.1: "three toy ships, one made of iron, the other of wood, and one out of loosely packed sugar, fall into a pool of water"

Sora 2; “a movie trailer whose genre you can never quite figure out because it keeps drawing on cliches from many different genres through a series of fast cuts, make it over the top, featuring movie guy voice”

So these researchers figured out you can basically hallucinate 3D cities into existence using just satellite photos & a diffusion model. The problem's pretty straightforward: satellites only see rooftops. Building facades? Invisible. Street-level detail? Doesn't exist. But people want flyable 3D environments, which means you need all that occluded geometry. When I worked on google maps photogrammetry, we could only use satellite-based 3D for isolated stuff like the pyramids - anything city-scale required airplane flyovers. Which is fine until you hit aerial-denied regions where you literally can't fly. Huge chunks of the world just unavailable. Their trick is honestly kind of beautiful. They train gaussian splats on satellite views, but as it descends toward ground level, the renders turn to absolute garbage - artifacts everywhere. Instead of fighting this, they just treat those nightmare renders as the input to a diffusion model. Basically - "hey FLUX, fix this mess." Then here's where it gets clever: they generate multiple diffusion samples per view instead of committing to one. Because any single denoising path is probably wrong in 3D space, but if you generate a couple and let the GS optimization find consensus across them, you get actual geometric consistency. They do this in episodes, curriculum style - start high, gradually descend (hence the name Skyfall-GS!). With each iteration the ground-level views get less fucked. By the end you've got real-time flyable cities that look surprisingly real, and the geometry still matches the satellite input. No 3D training data. No street-level photos. Just satellites + diffusion doing what it does best - filling in the blanks. It's like neural scene completion but actually practical, and it unlocks basically the entire world.

The new Sora Pro feature that builds storyboards and executes them is really interesting. Here is the prompt “an ad for the abstract concept of the feeling you get after falling asleep on your arm” Notice the high character consistency, multiple shots, composition. All the AI.

TLDR: OpenAI Atlas > Perplexity Comet in an agent mode head to head. Here is my use case: I have a very real, very tedious use case, which is a manual task that I do every day. 1. I go to the school website to look at each of my daughter's classes 2. I look at her grades 3. I look at her assignments, quizzes, all classwork that are due 4. I make a table that keeps track of all of this, which helps keep both of us accountable I asked both Comet and Atlas to help me with this task, since I have to do a lot of manual clicking, scrolling, reading, and data input. I scored them based on context (understanding the task), speed (how long it took), task (completing it correctly). Comet: Understood what I meant, made a few mistakes while clicking around, and seemed to take a while between thinking <> action, did not complete the task correctly (did not populate the table correctly). Atlas: Understood what I meant, was generally faster between thinking <> action and also did a lot of very human-like exploration of the website (clicked on 'missing assigments' to see if it was a good filter - it wasn't, and self-corrected), and went back to manually clicking through the classes, scrolling, and keeping track of what was on each page. It completed the table flawlessly (on a different page). Overall I still don't have a ton of agentic web browsing use cases, but it's great to be able to automate this one particular task!

Kramink is the guy in the movie In The Company of Men where the man won't stop talking to the deaf woman and the camera cuts to her point of view, complete silence, with this guy just moving his mouth yapping incessantly.

Stop being psychopathic memers and do your job effectively, fairly, and without emotion.

"Moments later, he [Naroditsky] mentioned 'the lingering effect of it' and became slightly choked up. 'One final game,' he said, his voice breaking, before signing off moments later." nytimes.com/2025/10/25/style…

Everyone watch this video. Kramnik is a sociopath. piped.video/gLVaiwl4lKg?si=d3_L…

Either allow replies or get off Twitter... what a poor setting to turn on for a CEO. You're muting the good along with the bad.

The service was beautiful. I will miss you always, Danya ❤️

And that's why there was so much stress on Danya. Non stop accusations from a former WC that influenced so many other people. The extent of this is way more than Hans whose accusation was ended within a year. To those who keep comparing Danya and Han, there's a lot of differences

Let me be clear, no amount of blaming Kramnik will ever be enough. It's just that we cannot let others sneakily get scot-free from accountability as well. People like Ian for example, who always fan the flames bts, as stated by Danya himself. Some Apologies are in order.