AI PANORAMA
The whole picture, any day you walk in.
Every other source of AI news is a stream. Miss a week and something that changed the world scrolls away below today's trivia, gone. This is not a stream. It is a state: whenever you walk in, you get the whole current condition of the field, laid out as a map you can fly through.
Free, open source, and every number carries a link to where it came from.
What is working right now
The first thing built was the hardest thing: real four dimensions. It now has real content in it — the first stories, written eight separate times over, by eight different AI models.
Hello, Tesseract — on this screen Mouse and keyboard. Works on any laptop or desktop. Hello, Tesseract — in Virtual Reality Meta Quest 3 and other WebXR headsets. Open this page in the headset's own browser, then press Enter VR.The first real results
5 stories so far, each one written 8 separate times over, by 8 different AI models. Every model read exactly the same frozen sources and did the entire job alone: the article, the one-line summary, the encyclopedia entries behind it, the tags, the links to other stories, and the instructions for its own illustration. Nothing any of them wrote has been corrected, improved or graded by us. If one of them is wrong, switch to another and see for yourself — that comparison is the point of this magazine, not a flaw in it.
Pick whose edition you want to fly through. Each one is its own four-dimensional world, because each model chose its own tags and its own links, and those choices are what decide where everything sits. Changing edition rearranges the sky.
| Edition | Cost per story | Words per article | Ideas explained | Links drawn | Links per node | Seconds each |
|---|---|---|---|---|---|---|
| Gemini 3.7 Flash | $0.0098 | 617 | 12 | 24 | 1.4 | 19 |
| Kimi K2.6 | $0.0483 | 700 | 12 | 25 | 1.5 | 195 |
| DeepSeek V4 Pro | $0.0533 | 788 | 15 | 37 | 1.9 | 166 |
| GPT-5.6 Terra | $0.0540 | 821 | 15 | 36 | 1.8 | 34 |
| Claude Sonnet 5 | $0.0809 | 555 | 10 | 25 | 1.7 | 51 |
| Grok 4.6 | $0.0865 | 807 | 16 | 38 | 1.8 | 467 |
| GLM 5.3 | $0.0990 | 953 | 19 | 59 | 2.5 | 235 |
| Qwen 3.8 Max | $0.1472 | 755 | 20 | 55 | 2.2 | 404 |
What the table already shows: GLM 5.3 wove a world 1.7 times more densely connected than Gemini 3.7 Flash’s, from the same 5 stories. GLM 5.3 wrote 953 words per article where Claude Sonnet 5 wrote 555. Grok 4.6 took 467 seconds to think where Gemini 3.7 Flash took 19. And Gemini 3.7 Flash, at $0.0098 a story, costs 15 times less than the dearest edition here — read them side by side and judge whether it shows.
Clicking a node inside a world does not open an article yet. The reading pages and the illustrations are the next thing being built.
What you are looking for
A tesseract is to a cube what a cube is to a square. A square has four corners, a cube has eight, a tesseract has sixteen. You cannot see one properly, any more than a drawing on paper can properly show a cube — but you can watch one turn, and that part your brain can genuinely learn.
Turn it in the ZW plane and watch it pass through itself and come out inside-out. Turn it four times and it comes exactly home. Nothing is broken when that happens; that is simply what four dimensions look like.
Why four dimensions, and not just a pretty effect
Because the fourth direction is going to mean something you can travel along. At one end hangs this week's raw, half-confirmed news. At the other end sits the settled encyclopedia. Swimming from one to the other is watching news condense into knowledge, with your own hand. Nowhere else on the internet can you do that, and it is the whole idea of the magazine made into a direction in space.
One promise about that: the fourth dimension will never be shown as a colour. Colour means identity here — which model, which edition — and nothing else.
Done already
- Real four dimensions, on a screen and in a headset, with five short lessons that teach anyone to follow an object through a turn it should not survive.
- Editions: every story written from scratch by eight different AI models, so you can judge them on work you actually came to read.
- Eight separate worlds, one per edition, because each model decided for itself what belongs beside what.
Coming, in this order
- The reading pages, with every source linked beside every number, every date and every name.
- The illustrations. Every model wrote instructions for its own picture, and the pictures get drawn with the same tool and the same starting point for all of them — so the only difference you see is how well each one directed the artist.
- Model comparisons in real dimensions. Not our own invented scores: the benchmarks the world already publishes, combined into three and four dimensions instead of the one-dimensional bar chart everybody else prints.
- A do-it-yourself workshop, the way the old computer magazines had. The first project is Night Watch.
Comfort, if you use the headset
The room never moves and you are never moved. Only the object turns, and it turns slowly on purpose. If anything is uncomfortable for even a moment, that is a fault worth reporting, not something to push through.