Have you ever wanted to see what Magic Eye images look like, but have never been able to catch the right focus? If you have a VR headset, I have a solution for you!

I vibecoded a page which sends different parts of the image to different eyes, at the right offset (it's calculated automatically), allowing you to see the stereo effect!

https://optozorax.github.io/stereo_viewer/

I suggest going very close to the wall to see stereo effect without flickering edges (or without flickering legs on the picture 5), and then slowly go back.

Turns out my whole life I focused on these images wrong! I always crossed my eyes, focusing on a point between me and the monitor, but this creates an inverted stereo effect, which isn't what these images are designed for. They are designed to focus on a point farther away than your monitor. I'd never have figured this out without that app! Now I can catch the right focus if I try hard enough.

I hope that when Steam Frame is out, I will be able to create an app that tracks your eyes and teaches you how to position them correctly for such an image for making muscle memory.

Just found a real Mandelbrot fractal happily flourishing in its natural habitat... 😁

#nature #mathematics #fractal #mandelbrot

There was a question on twitter about what happens if you go through a portal with one curved part. Here are the simulations.

In the first one we have a very rigid object going through such portal with a little push. It's almost fine in the end.

In the second one we have a very flexible jelly-like object going through very very curved portal. It's also fine.

But in the third one we have very rigid object going through very curved portal. It can't go by itself, so we have to push it with a wall, and it just rips apart...

Don't go through strange portals, guys...

The way it works is just a scanner-printer portal (I described this concept in my most popular video) with different entrance and different exit. It violates the surface irrelevance axiom, because with different surface it acts differently, so the transformation that it does is not equivalent to flat space. You would feel with your body when you are going through such portal. And the image would be distorted too, like through some lens.

Some people propose to call them "portal lenses". Maybe I will make a video about them one day with all these animations and simulations.

What else would you like to see with such portals?

A while back, I wanted to build a diff algorithm capable of detecting moved blocks of code (which is super handy for refactoring). I slowly started exploring the topic and vibecoding a prototype. I decided to start by implementing manual control for a standard diff algorithm, so no move detection just yet. Here is the link:

https://optozorax.github.io/hand_diff/

You can tell the diff algorithm exactly how to interpret each line: whether it should be considered deleted, added, or unchanged. You can do this using the small +/-/= buttons next to each line. The diff recalculates immediately upon clicking.

There is already an example showing where standard diff algorithms produce a messy result (image 1). It looks like `.someblock` extended, closed inside itself, and then `.someotherblock` started. It's honestly painful to look at.

Now, here's how to fix it in the editor using two methods: click the plus sign on line 19 (right side) or the equals sign on line 12 (right side) (image 2). You can see that after manual tagging, the diff looks much more logical: a property was added to `.someblock` and a new `.someotherblock` was added. Before, it wasn't clear that `.someblock` had changed - the diff was misleading. Now, it's objectively better.

It looks promising so far. I hope it works just as well when I add move detection.

Everyone says that space in games like Pacman or Asteroids is topologically a torus, because its upper edge is glued to its lower edge, and the same for left/right.

But is it true geometrically? What if you visualize the space of a 3D torus where light rays follow geodesics? Turns out, it will not look as you expect.

And also, how would 2D wormholes look?

All of this is in this video of my friend. I also slightly participated in the creation of this video (we programmed the underlying visualization engine together).

https://youtu.be/9JbYDDjhnx4

Since I already have a Finite Element Method program for calculating gravity through portals, I figured I had to render Bad Apple with it.

https://www.youtube.com/watch?v=Zaq-aWlAgFk

In my previous video, a lot of people asked how gravity would change if you rotate one of the portals.

This video shows what happens in two cases: when the portals are at different heights, and when they're at the same height.

Do portals violate conservation of energy? No! You just need to teleport gravity through them. How do you do that? And what does the Finite Element Method have to do with it? Watch it in my new video 🟠🔵.

This is the best thing I've ever done in my life: a 40-minute video, 5 months of work, 230 hours of programming.

https://youtu.be/DydIhwLrbMk

120-cell (4D figure)

New video! Portalology keeps going deeper and deeper.

"How regular portals create infinite recursive space"

https://youtu.be/1UsAqvooYj4

🔵🟠 New portal video!!! 🍿

It's about the most incredible discovery about portals. It predicts pocket dimension, and flips our understanding of the portal physics!

https://youtu.be/IhEaw3Kuhf0

I made a visualization of how portals bend space.

https://optozorax.github.io/space_bending/ - works even on phones

The second video (in the comments) shows a flat space with two double-sided portals.

The first video shows how these portals bend space. You can see that for the blue square, it creates a local cylinder, so it's basically going in circles and its space is locally looped. For the green square, the space just became adjacent, as if there's no distance between its parts. Very interesting shape, I didn't expect that - reminds me of a wormhole shape, if you glue the singularity points together in a special way. Looks like I've finally concluded that portals are wormholes and nothing else.

This shape is created through physical simulation with particles and springs between particles, and angular springs (which tend toward a specific angle). I integrate forces through 4th-order Runge-Kutta, since naive integration was too unstable and oscillating. Most of the initial code was written by Claude 3.7 :)

The result is super interesting, I learned a lot of new stuff. In the future, I want to experiment with how space bending looks for portal-in-portal, scaling portals, and other stuff. I'll need to make a separate YouTube video about this too, but first I need to finish the current one...

I published a new video about strange portals! Now it's portals from Mobius strip, trefoil knot and etc.

https://www.youtube.com/watch?v=3z3aC-1MGoE

Someone proposed skew monoportal in comments, and that's really works!!!

I made a new video about portals!

This one is about portals with three parts, portals with one part (?), and other interesting portals that I discovered.

https://youtu.be/DTcfaHfDCEc

I finally finished my first youtube video about portals :)

I show what will happen if you put one portal into another, how it works, and many more interesting related things. For this video I wrote a program that simulates and renders this.

https://www.youtube.com/watch?v=1zgBwQVr-wI

So, I just translated, improved and dubbed to English (with my voice) my first video about portals. Gonna create English channel. Does someone English-native wants to review it?

(This all is a translation of my recent posts from my russian telegram channel. I will publish more here after I continue working on this project. Someday I want to make a couple of YouTube videos about this topic, in Russian and English.)

In general I'm very very happy with the new paradigm, feel like some kind of pioneer, because I've never seen this for agent training tasks! Now I'll try to formulate everything as energy-based model, maybe training will be much simpler! If you have ideas what approaches to try - write!

PS: This is why I rejected your stupid Reinforcement Learning at the very beginning and chose genetic algorithms, because can you imagine me conducting the same experiment with RL? I would never do it, I wouldn't have enough brains, time, or computer to train Energy-based model through RL. That's why they haven't earned such popularity, because they're damn hard to train. But for genetic algorithm I coded this in half an hour, and during training it couldn't care less. This is called FREEDOM.

#ai_car