Design and Code @ Oxide. Formerly Pentagram
London
- Tweets 1,093
- Following 645
- Followers 2,434
- Likes 4,795
Pinned Tweet
Web 3d explorer for @oxidecomputer. Built in three.js/r3f.
Tonnes of little interactions and details in this. Poke around, and take a look at the guided tours.
explorer.oxide.computer
34
166
1,776
102,609
Ben Leonard retweeted
Absolute stunners from the California Department of Water Resources on hydro infrastructure. Give every state project a Photography Unit
15
236
3,212
126,539
Ben Leonard retweeted
does Jev understand colour?
incredible model, very cheap and fast. definitely feels like a leap forward for creating new UX/UI paradigms on top of modern language models.
100
355
5,730
540,214
Ben Leonard retweeted
Navajo weaver Marilou Schultz (1954-) and her woven chips
52
4,407
39,211
1,014,839
Ben Leonard retweeted
Three.js conf time! Look how amazing this speaker badge is!
It has games!!
72
107
1,568
100,338
Only a few hours to go before the launch of tldraw flash, our new animation tool. @steveruizok strapped in to show how the app lets you perform animations by hand
10
7
212
12,488
Ben Leonard retweeted
“Meltdown’s” own source code running and melting on the four pillars of the legendary SEIBU building in Shibuya.
February 2026
10
16
142
4,977
Ben Leonard retweeted
Mini Pi plus successfully completed the long jump, taking off and landing smoothly!
#WorldHumanoidRobotGames
#Cuterobot #Opensourcerobot
25
208
1,730
278,992
Ben Leonard retweeted
There is also the opposite in type design, that is extreme analytical approaches. Here is our custom Nelder-Mead solver that proposes Houston Mono chamfer nodes in ~2000 iterations. The goal is to smoothly transition from one direction to another using exactly 6 nodes with a 'smooth chamfer', i.e. two rounded 45 deg chamfer corners.
Basically, I care about 3 components for Houston Mono chamfers:
1) How the stroke width smoothly changes from one thickness to another. Here the vertical stroke is 84. It transitions to horizontal at 80. But the midpoint of the chamfer is like ~85.63 (see the table)! It's not linear, but the solver makes this as even as possible.
2) How each of the 45 deg turns curve, i.e. curve tension. Held at 91% by the solver. That means the handles span 91% to the Tunni point (where two handles intersect).
3) How deep the chamfer is...O has deeper chamfers than 2 for example.
There are a few other things like I can bias/skew the stroke change as I see fit visually.
In type design, there are all these tools to help designers. For e.g., you can predefine specific stems (or autodetect) and set their values. So what happens is that the designer now follows these rigid rules and if they're violated, it's exactly like a linter/type checker. Errors are displayed on the screen. The designer tends to correct them, but soooo many times these rules just don't apply. It's a trap. The solution is to keep a printed sheet on your desk with ideal values and unliked from the software. Then nudge them as needed. Auto-hinting will suffer slightly but in vast majority of cases there is a higher priority of making the glyph look consistent with the rest of the typeface.
ALT Source: https://www.reddit.com/r/FontLab/comments/1mo0edh/fontlab_8_tip_stems/
11
15
399
23,050
Some artwork prep for upcoming event. I've always had the habit of choosing colours that cannot be mistaken for the final artwork to avoid print mistakes.
2
483
Added motion character mapping to my ASCII tool inspired by @MaximeHeckel fantastic post "Shading Motion" blog.maximeheckel.com/posts/…
5
1
65
2,250
Playing with some GIF experiments done a long time ago by @luisgurico
1
8
383
Ben Leonard retweeted
New post 📝
For this article, I asked myself: what happens when motion becomes an input to our shaders?
To answer it, I spent this summer experimenting with effects like motion masking, blob tracking, optical flow, and motion blur
blog.maximeheckel.com/posts/…
ALT Shading Motion
24
62
638
82,244
