A browser tab you can get lost in.
I've been expanding Orrery with @OpenAI 's Astra: five scales, your night sky, Orion in 3D, exoplanets, science experiments and a real JPL flyby.
Pick a world. Start exploring.
https://t.co/jyhYEb7jZN
ChatGPT’s image capability is seriously underrated and undersold. For Orrery, I used it to think visually and generate the early concepts. Codex’s Sol model then relentlessly turned those ideas into a working product.
https://t.co/bw88KEFcNS
The feet aren't animated, they're placed.
Claude wrote a footstep planner that picks where each foot lands before the walk starts. A planted foot is pinned to that spot until its next step, so it can't slide. The hip, knees and legs are then solved backwards from the feet with 2-bone IK.
Then it measured every planted foot across the whole film. Max drift: 0.01 px 🏮
Made a music video with Claude Opus 5.5 🏮
No AI video generator. Every frame is drawn by Python code Claude wrote: a rice-grain + paper-cut stop-motion world, a puppet rig that actually walks, and lyrics synced word by word.
Song: @suno
Direction: me
Everything else: @claudeai
Whoomp there it is!
After two astronomers, Urbain Le Verrier and John Couch Adams, independently calculated its predicted location, Johann Gottfried Galle found the planet Neptune 180 years ago on Sept 23, 1846.
In 1989 when Voyager 2 became the first—and so far only—spacecraft to visit Neptune, it took thousands of images of Neptune, its rings, and its moons, including this one.
Quick caveat: these are AI-generated images, so they may not be technically accurate. I am sharing this mainly to show the jump in image quality and visual understanding between Image 2 and 2.5.
ChatGPT Image 2 → Image 2.5 is a serious jump.
Same task. Same AirTag.
Image 2 made something that looked convincing.
Image 2.5 did a much better job explaining how it actually works. That’s the leap:
image generation → visual reasoning.
Side-by-side 👇
ChatGPT Images 2.5—faster, sharper, smarter, with better tools for creating whatever you can dream of.
- Faster image generation to keep your ideas flowing
- Improved fidelity for more natural, recognizable images
- Consistent details across multiple edits
- Comment-based edits to change only what you want
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
Try it yourself 🚀
Orbit Academy turns math practice into an adventure — where even mistakes become part of the journey.
https://t.co/XlRFgFinzC
Built with Astra.
What if a wrong answer started an adventure?
Orbit Academy turns mistakes into Mistake Monsters-then lets kids learn, rematch, and keep flying.
Tap-to-blast math, visual hints, number gates, and custom math, physics & chemistry missions.
Built with @OpenAI’s Astra in Codex. WIP🚀
What if you could trace any technology back to the Earth?
Material Earth maps the minerals behind batteries, motors, aerospace and more—connecting designs to production, reserves, resources and geopolitical dependencies.
Built with @OpenAI 's Astra. 🌍
https://t.co/sxX4jiM9Iu
Asked Astra to make a free classics library feel alive. It delivered.
Turning pages in Three.js. Book spines that lift as you explore. A quiet place to read.
The little details make Open Stacks (https://t.co/WtzTsaAV0o). Credit @OpenAI 's Astra for the craft.
Come wander 📚
What if a wrong answer started an adventure?
Orbit Academy turns mistakes into Mistake Monsters-then lets kids learn, rematch, and keep flying.
Tap-to-blast math, visual hints, number gates, and custom math, physics & chemistry missions.
Built with @OpenAI’s Astra in Codex. WIP🚀
inspired by @skirano’s blender demo, had to try @OpenAI 's Astra myself!!
gave it a reference and asked for a music sampler + an exploded-view animation. it wrote the python, checked a preview, refined it, and delivered editable blender files.
very impressed with this!!
I had early access to GPT-6. This model will completely shatter your understanding of what's possible, and it will start a new era of creativity.
It's hard to explain how big of a jump this is, so I'll share my tests.
1/6 Made this 3D model and animation in code from an image.
The 400 m was even crazier... 🤯
TianGong (Tien Kung) took the win in 38.15 seconds, leaving Honor’s three Lightning humanoids far behind.
A little payback for the robot marathon too.
And these races are fully autonomous now. Once they’re on the track, it’s completely up to the robots.
The language we choose reflects our personality and the kind of person we are. Choose your words wisely, because over time, you become what you think, what you speak, and what you practice.
Using bad language is not cool. Abusing is not cool.
Let’s keep the definition of cool simple: being hardworking, kind, respectful towards others and yourself, striving to be successful through honesty, and making your parents and your country proud.
THAT’S COOL.
I’m speaking about this because, over the past few years, the way we perceive abusive language has changed. For many, using such words has become so normal that it’s even considered “cool.” Without realizing it, people casually use degrading and disrespectful language in everyday conversations.