Penrose retweeted
Replying to @OpenAI
@OpenAI's new 4o image generation is pretty amazing.
It also fails on some utterly basic tasks.
Maybe a good metaphor for how easily we can be misled by algorithms based largely on fitting empirical data (i.e., machine learning) rather than deductive reasoning. @UsePenrose
Penrose retweeted
talked about diagrams and defended my thing today 🎓 thanks for friends who joined in person & on zoom!
Penrose retweeted
video of the defense talk is up! youtu.be/gVZJc7_0T14?si=fJhP…
Penrose retweeted
took the idea of “run the layout optimizer again when interacting with the diagram” pretty far this summer. Bloom is a React library for interactive diagrams. It lets you reuse the diagram styling just like Penrose, but in a general-purpose programming language.
Penrose retweeted
Great work by the @UsePenrose team building Bloom: a lightweight way to make interactive diagrams: penrose.cs.cmu.edu/blog/bloo…
All coordinates are automatically figured out by the Penrose layout engine, and diagram specifications can be re-used for different content (like HTML/CSS).
Penrose retweeted
A standard way to draw planar graphs is to draw nodes as dots and edges as line segments.
But the Koebe–Andreev–Thurston theorem also shows that any planar graph can be visualized by drawing nodes as circular disks, which are tangent if they share an edge:
Penrose retweeted
In a moment nobody was waiting for, I've released meshes from Mark Kilgard's classic OpenGL / GLUT demo "dinoshade.c": cs.cmu.edu/~kmcrane/Projects…
The image below re-imagines this example as a vectorized SVG generated in @UsePenrose—read how it was done here! github.com/penrose/penrose/t…
Penrose retweeted
penrose.cs.cmu.edu/
Penrose (now released) is promising looking domain-specific language for scientific figures that isn't bogged down by decades of tech-debt (cf Tikz/tex)
Esp like the .style/.substance/.domain division, makes iteration easy (e.g. shape in 2, 'Pipe' in 3)
Penrose retweeted
Tired of making the same kind of diagrams over and over by hand (e.g., in PowerPoint)?
The @UsePenrose team has been working away on Penrose 3.0, an automated notation-to-diagram tool, finally released today!
Check it out here: penrose.cs.cmu.edu/blog/v3
Penrose retweeted
The winding number of a polygon is, quite literally, the number of times a polygon "winds" around a given point x.
It can be used for inside/outside tests by just summing the angles made by each (oriented) edge: zero if x is outside, ±2π if x is inside.
[Made with @UsePenrose.]
Penrose retweeted
Animation of how the winding number changes as the center point x moves inside/outside a shape.
Notice how the signed angle of an edge suddenly jumps from +π to -π (or vice versa) as x crosses an edge, changing the total signed angle by ±2π.
[Made with @UsePenrose]
The winding number of a polygon is, quite literally, the number of times a polygon "winds" around a given point x.
It can be used for inside/outside tests by just summing the angles made by each (oriented) edge: zero if x is outside, ±2π if x is inside.
[Made with @UsePenrose.]
Penrose retweeted
Thanks to @sgestep, we squashed a 4+ year old bug in @UsePenrose that makes it *asymptotically* faster! 🪲
Pro tip: if you're using L-BFGS to avoid building a dense matrix… don't build the dense matrix!!
Try the new, way faster version of Penrose here:
penrose.cs.cmu.edu/try/
Penrose retweeted
What's the best way to pack N circles into a square?
Using less than 20 lines of code in @UsePenrose, we can reproduce the best-known solutions ever found: en.wikipedia.org/wiki/Circle…
Try it out for yourself!
penrose.cs.cmu.edu/try/?gist…
Penrose retweeted
You can perfectly pack 16 unit squares into a larger 4x4 square.
But what’s the smallest square that can contain 17 squares?
Check out this thread for an interactive exploration, via @UsePenrose!
Replying to @keenanisalive
[19/n] And keep on going—how low can we go?
Warning: if you make it too small, the optimization may go on forever and freeze your tab! Sorry about that.
You can also slow down the optimization to a step size of 1 to see how it works:
Penrose retweeted
[1/n] There's been a lot of hubbub lately about the best known packing of 17 unit squares into a larger square, owing to this post: xitter.cf/KangarooPhysics/status…
I realized this can be coded up in < 5 minutes in the browser via @UsePenrose, and gave it a try. Pretty darn close! 🧵
Penrose retweeted
Happy birthday to Don Knuth! Knuth received the 1974 #ACMTuringAward for his major contributions to the analysis of #algorithms and the design of #programming languages. Watch this video where Knuth discusses the origins & motivations of the TeX project: bit.ly/3IVls6D
Penrose retweeted
A pool player who works on @UsePenrose must do Made with Penrose™ pool diagrams. Gave a talk about pool yesterday in the @S3DatCMU weekly seminar to show how a pool player plans their shots.
Penrose retweeted
Little experiment with direct editing in @UsePenrose.
The beauty of Penrose is that it automatically generates diagrams for you, from simple CSS-like code. But what if you want to make some small tweaks—while respecting constraints?
We're working on it…
penrose.cs.cmu.edu/siggraph2…
Penrose retweeted
Correct typography is something you notice only when it *doesn't* work!
Lesson of the day: to draw a bounding box around text (assuming an alphabetic baseline), you must subtract half the bounding box height, then add the descender height.
Penrose retweeted
Another fun little @UsePenrose example: exterior algebra: github.com/penrose/penrose/t…
I just implemented a basic "wedge" operation on vectors—the image shows some random example diagrams generated from the displayed Substance code. Can you take it further? :-)





















