a model where you can mark regions of an image to remove and the model imagines what should fill the space. The released model required PyTorch and NVIDIA CUDA, but since it described itself as 0.2B I decided to try and get it running using WebGPU in a browser. TL;DR: I got it working, and you can try the demo at simonw.github.io/moebius-web/. Read on for the details.

Outsail the Royal Navy in your browser. Real sailing physics, 5-minute voyages, 30 historical treasures to recover. Free to play.

I thought about this the other day, and I thought it’d be fun to share this internal tool I made over a decade ago to aid with exploring options for Medium’s typographical redesign.

It’s called Fontificator. You can play with Fontificator here (desktop browsers only), or watch the likely confusing video below:

The motivation for building Fontificator came from two observations:

font previews on type foundry sites were generally too limited to get a real sense of how a certain typeface feels, and it was best to see a font in situ,
often an extremely tiny nuance – like adding some letter spacing, or messing with line height – was what separated something that was promising from something that seemed very far from working.
With Fontificator, I was aiming at this Doug Engelbart-esque notion of one hand on the keyboard + one hand on the mouse, and the UI where it was only necessary to point to an element, and the keys under your other hand would start working immediately – no clicking needed:

F and G to change the font,
– and + for font size,
← and → for letter spacing,
↑ and ↓ for line height,
< and > for opacity (for all the above you can hold Shift for bigger moves),
and, there are a few more shortcuts you can see at the top.
This way, we could move really, really fast. To accommodate that, Fontificator always tried to keep the current item under the cursor by counter-adjusting scroll position as needed.

On top of it all, a few more shortcuts:

⇥ and ⇧⇥ move very quickly between different types of stories so you can preview that,
Space compares to the original/​current version,
1–9 allow you to switch to different “slots” so you can have various presets ready to compare,
Esc hides the toolbar for maximum immersion,
⇧R resets.
You can also edit any text if you are so inclined, and also drag in any font file from your computer onto a paragraph – then that font becomes part of the F/G stack. (Bernino Sans and Freight Text were the starting fonts before the redesign.) On the left, you can also see a naïve mobile preview – there was also more sophisticated on-smartphone preview, but I removed it from this restored version.

A real-time collaborative whiteboard for brainstorming, teaching, and sketching ideas together. Scribble on your iPad or laptop browser and see it instantly sync across devices.

Collaborative Note Editor

A simple, blazingly fast, self-hosted URL shortener with no unnecessary features; written in Rust.

The Uncertainty Calculator enables users to perform calculations using precise numbers and ranges (e.g., 4 or 4 ± 6) to express uncertainty. It simplifies statistical reasoning for everyday decisions, helping users gauge potential financial outcomes in uncertain situations.

Key Features

• Input both exact numbers and ranges with a tilde (~) to denote uncertainty.

• Operates on basic mathematical operations: addition, subtraction, multiplication, and division, and supports functions like exponentiation, square roots, and trigonometric functions.

• Utilizes a Monte Carlo method for computations, providing a probabilistic range of results.

• Design is basic and intentionally user-friendly, aiming to empower those without deep statistical knowledge.

Limitations

• Calculation speed is slower due to Monte Carlo simulations, and the formula parser may show errors.

• Only normal distributions are assumed for ranges, providing a simplified model for uncertainty.

Practical Applications

• Estimate financial viability of business ideas, calculate potential investment returns, assess risks, and more, all accounting for uncertainty in key variables.

Hello HN! We’re a small creative studio specializing in real-time 3D experiences. Netlify approached us to design and build an interactive experience to celebrate reaching 5 million developers.
Inspired by the classic game Marble Madness, we created a gamified experience where users control a ball through playful, interactive levels. The goal was to blend marketing content with the look and feel of a game to engage users.

The app is built with Three.js [1], using our custom render pipeline and shaders, and uses Rapier for