Pinned Tweet
AI was supposed to save us time.
Now we can’t stop watching it work.
Orac, our @ChatGPT agent, navigates tax portals, plans trips & builds apps like a gamer clearing impossible levels.
It feels like Twitch.
Except the final boss is bureaucracy.
And the loot is our lives back.
AI isn’t just explaining science. It may be advancing its foundations.
OpenAI claims breakthroughs in computing, wave mathematics and quantum physics.
Still under expert scrutiny. Potentially historic if confirmed.
Four posters explaining what’s claimed—and why it matters. 👇
Stay curious. There’s always more to discover.
Remarkable ideas. Hidden connections. New perspectives.
Explore SparkZone - short, visual stories that help you see the world differently.
sparkzone.flon.co
🤖 Made with AI
OpenAI model names aren't that confusing, you just need to know we started with GPT-1, 2 and 3, then 3.5, then 3.5 Turbo, then GPT-4, then GPT-4 Turbo, then GPT-4o where the o means omni, except it didn't really end up being omni, then GPT-4o mini
Then we switch to the o series, where the o now means OpenAI instead of omni, starting with o1-preview and o1-mini, then o1 drops the preview, then we skip o2 and they announce o3 and o3-mini, but o3-mini gets released first, meanwhile GPT-4.5 comes out and THEN GPT-4.1 because why not, alongside GPT-4.1 mini and nano, then o3 finally arrives with o4-mini even though there's no regular o4
Then we're back to GPT with GPT-5, mini and nano, followed by 5.1, 5.2 and 5.2-Codex, then 5.3-Codex, and yes there's also GPT-5.3 but ONLY Instant, no Thinking or Pro, then 5.4 brings the coding capabilities into the main model so we drop the Codex suffix, but hey mini and nano are back, then 5.5, then 5.6 where the names are now Sol, Terra and Luna
Then GPT-6 arrives but only Astra at first, followed by Sol and Luna, no GPT-6 Terra this time, and exactly a week after GPT-6 Sol we get GPT-6.1 Sol with near-Astra performance at a much lower cost
And that's just the quick version, wasn't that hard was it?
FLON retweeted
I set a camera really close to the launch pad to get this shot. It was a sound activated trigger since it wasn’t safe for me to stand here.
That’s all 33 Raptor engines performing nominally as Starship heads to orbit for the first time.
FLON retweeted
2022: KI kann nichtmal 2+5 korrekt rechnen. 😂😂
2023: KI kann nichtmal Buchstaben in Wörtern zählen. 😅
2024: KI kann nichtmal Bilder eines randvollen Weinglases generieren 🤓
2025: KI will zu Fuß mit dem Auto in die Waschstraße 🤦♂️
2026: KI kann Millenium Mathe Probleme nur durch Bruteforce lösen😖
2027: 👀
No, no, you’re missing the important bit - the 10,000-agent swarm didn’t just solve Navier–Stokes. It changes the economics of scientific discovery!
Tao says we’re optimizing the wrong objective function, but if we use Specification Agents to identify reusable mathematical keys …
Got a 10,000-agent AI research swarm?
Don’t waste it collecting famous mathematical trophies.
Ask scientists for the problems that would unlock entire fields.
The best breakthrough isn’t another door opened.
It’s a new key that opens 1,000 doors. 🔑
Physicists. Mathematicians. Biologists. Chemists. Materials scientists…
What should we aim the swarm at?
🤖 Made with AI
I find these World Trade Center simulations by Kostack Studio fascinating because they remove the smoke and let you see the collapse much more clearly.
In the real footage, so much is hidden by fire, dust, debris, distance, and chaos that many people come away with the impression that the towers simply dropped straight down into their own footprints.
But that is not really what happened.
These simulations make it much easier to see that the collapse begins in the heavily damaged and fire-weakened region, the upper section starts moving and tilting, and the destruction then propagates downward through the structure.
The towers do not simply disappear vertically into their own footprints. Large portions of the upper structure and exterior are moving outward and to the side as the collapse develops. Much of that motion is simply obscured in the real videos once the enormous debris cloud forms.
And this is also why I find these simulations useful when people describe the collapses as looking like “controlled demolition.”
Strip away the smoke and dust and the resemblance becomes much less convincing.
What you see is a highly asymmetric structural failure beginning around the aircraft-damaged and fire-affected floors, followed by an enormous moving mass progressively overwhelming the structure beneath it. There is tilting, rotation, exterior columns peeling away, debris being thrown far outside the building footprint, and structural failure occurring chaotically as the collapse moves downward.
That is very different from the neat mental image usually implied by “controlled demolition,” where structural supports are deliberately severed in a carefully timed sequence to produce a predetermined collapse.
And importantly, we already know what preceded these collapses.
Both towers suffered massive aircraft-impact damage. Fireproofing was stripped from portions of the structure. Large, uncontrolled fires then burned across multiple floors - for about 56 minutes in the South Tower and about 102 minutes in the North Tower - heating and weakening structural components until collapse initiated.
Once that upper mass began moving, the loads imposed on the floors and structure beneath became vastly greater than anything they had been designed to carry.
The result was progressive collapse.
No, an animation is not proof simply because it looks convincing, and these are not official NIST simulations. They are visualizations built under particular assumptions.
But as visualizations, I think they are extremely useful.
They let you see something the original footage often cannot show clearly: beneath that enormous cloud of smoke and debris was a complicated, violent, asymmetric structural failure - not two buildings neatly dropping straight down into their own footprints.
A note on the simulation: this is not intended to reproduce the collapse down to the exact motion of every beam, floor panel, and fragment of debris. No practical simulation could do that. Its value is in modeling the large-scale structural behavior - how the major sections deform, move, tilt, separate, and progressively fail. Think of it as a visualization of the mechanics and overall collapse behavior, not a frame-by-frame reconstruction of where every piece of steel ultimately landed.



























