Visualizations
A collection of links I’ve found that provide a nice visualization of some topic or concept.
2025-04-29 — Another float format visualization tool
https://mserdarsanli.github.io/FloatInfo/ — this link is more purely a visualization tool, not a blog post like the link in the section below.
2025-04-28 — Chrome comic
https://www.google.com/googlebooks/chrome/big_00.html — an oldie but a goodie, this comic explains the motivations behind building the Chrome web browser, and what Chrome did differently.
2025-04-21 — Floating point numbers
https://dennisforbes.ca/articles/understanding-floating-point-numbers.html — a short and sweet blog explaining the various different floating point number binary formats that are used in modern machine learning.
2025-04-07 — Over-engineered anchor links
https://thirty-five.com/overengineered-anchoring — a somewhat nerdy deep-dive into the problem of anchor link scroll behavior, with some nice visualizations explaining the problem.
2025-03-22 — Autogression vs diffusion vs block diffusion
https://www.reddit.com/r/singularity/comments/1jbdm3t/block_diffusion/ — “traditional” LLMs use autogression to ‘predict the next token’ one token at a time. Recently, there’s been some buzz about text diffusion models that work more like the diffusion image generation models. Instead of generating one token at a time, they produce the entire image in one shot by organizing it incrementally out of noise. Applying that to text is an interesting idea.
Block diffusion seems to be a combination technique, producing incremental chucks using diffusion.
2025-03-21 — Matrix multiplication
https://pytorch.org/blog/inside-the-matrix/ — a great 3D tool for visualizing the high-level dynamics of a large matrix multiplication, like that inside of the attention mechanism of an LLM.
2024-12-11 — Positional encoding
https://fleetwood.dev/posts/you-could-have-designed-SOTA-positional-encoding — visualization of the positional encoding algorithm used by transformer LLMs.
2024-12-10 — Interactive data structure visualizations
https://www.cs.usfca.edu/~galles/visualization/Algorithms.html — interactive visualizer tools for a bunch of common (and not so common) algorithms. The Recursive Reverse example is a simple one to start playing with.
2024-12-02 — Visual information theory
https://colah.github.io/posts/2015-09-Visual-Information/ — in-depth explainer of information theory, including visualizations of probabilities and how they compose and interact.
2024-11-20 — Complex numbers and Anti-aliasing
https://acko.net/blog/how-to-fold-a-julia-fractal/ — while not the main point of the post, this article contains a fantasticly intuitive visual explanation of complex numbers.
Along with Tau, a FANTASTIC example of how improving on how something is first taught can make something go from completely opaque to very intuitive.
The point about how 1<90 degrees is what we call “i”, and how we write complex numbers as confusingly as an X+Y coordinate instead of magnitude @ angle was eye opening.
This reminds me of the flash of insight that followed from reading the Tau manifesto and realizing that thinking of Tau as “one turn” makes angles and their trigonometric relations suddenly comprehensible.
https://blog.frost.kiwi/analytical-anti-aliasing/ — visual explainer of the problem anti-aliasing algorithms solve.
2024-10-08 — LLM tokenizer
https://tiktokenizer.vercel.app/ — visualizes how a piece of text is tokenized for being fed to an LLM.
2024-10-05 — The Data Visualization Catalogue
https://datavizcatalogue.com/index.html — an external (meta) collection documenting many different kinds of visualizations, and software that can be used to make that kind of data visualization.
2024-08-13 — UTF-8 Visualizer
https://sonarsource.github.io/utf8-visualizer/ — visualize how the bytes in a UTF-8 string are encoded. The user can input their own string to experiment.
2024-08-09 — LLM ‘Transformer’
https://poloclub.github.io/transformer-explainer/ — visualization of how the transformer block at the core of large language model (LLM) systems works.
2024-07-21 — Locations of vehicle accidents
https://roadway.report/testmap — interactive map showing vehicle accidents in a chosen area.
If you click on one of the data points and look at all the fields that are filed out for each accident, its an interesting and impressive example of data fields that are specific enough to be queryable in a structured way.
Impressive in both its design, and apparently in how carefully each report must need to be filled out by the officers on scene.
A great example of the complexity of the modern world.
2024-07-11 — 2D planes in 3D space
https://alexharri.com/blog/planes — visualizations about math involving 2D planes in 3D space
“Learning about planes felt abstract and non-intuitive to me. “Sure, that’s a plane equation, but what do I do with it? What does a plane look like?” It took some time for me to build an intuition for how to reason about and work with them. In writing this, I want to provide you with an introduction that focuses on building a practical, intuitive understanding of planes. I hope to achieve this through the use of visual (and interactive!) explanations which will accompany us as we work through progressively more complex problems.”