Josh On Design

I find Large Language Models fascinating. They are a very different approach to AI than most of the 60 years of AI research and show great promise. At the same time they are just technology. They aren’t magic. They aren’t even very good technology yet. LLM hype has vastly outpaced reality and I think we are due for a correction, possibly even a bubble pop. Furthermore, I think future AI progress is going to happen on the app / UX side, not on the core models, which are already starting to show their scaling limits. Let’s dig in. Better pour a cup of coffee. This could be a long one.

Working in the Mixed Reality group at Mozilla was the most fun I ever had in my 30 year career. Helping to usher a new technology into the world. One with such potential. I’d really love to go back there. Unfortunately it seems there isn’t really a there to go back to. Cue memory fade.

I’ve been quiet lately, not because I’ve had nothing to say but because I’ve not felt well enough to say it. That changes today. 

One of my original IdealOS blog posts from 2017 showed up on the front page of Hackernews the other day (comments here). This got me thinking about IdealOS again. I haven’t worked on it in a couple of years, but as I read through the comments and links to articles by people with similar ideas, I came to a realization. I am still working on it. Maybe not directly, but I’m still exploring ideas that are needed to build IdealOS. So with that in mind let’s take a look at what I’ve been working on lately.

It's been a bit since I've posted on my round screen watch project. Most of my time has been taken up by work, travel, and family stuff, but I did have a few seconds to add a feature or two.

It’s a truism of the web that when something becomes free it turns to crap. This is largely due to advertising, tracking, and SEO hackers getting their crappy version of something to the top of the search index. When this has happened to a tool I want to use I can’t fix the underlying problem, but I can make my own non-tracking version and share it with the web, doing my little part to make the world a better place. Today it’s a CSS Text Style builder.

I’m continuing to work on a little smartwatch prototype using a little round LCD and I want to have nice looking text. The default font for CircuitPython is fine, but it’s very tiny. This LCD has a pretty high DPI compared to other hobbyist screens( > 200 ppi), so I need to find a new font. CP has a way of importing new fonts, but there are a few pitfalls and tricks I discovered, so that’s what I’m covering today.

For my next project with the WaveShare Round LCD I want to create an animation of Dr Strange's Eye of Agamotto. Here is my first attempt in JS (looks more like the comics than the movies). Not bad, I think.

I'm enamored of all of cheap and hackable screens and embedded CPUs coming onto the market. I'm frustrated, however, that despite embedded computing being more accessible than ever, the mass manufacturers products are all becoming more and more the same. Where are the phones with oval screens? Where are the steam punk pocket calculators? We have the technology to make devices that are ever more niche and increasingly creative and different, yet what you can actually buy is more glowing rectangles. I'd like to fix this.

I love HTML Canvas. I even wrote a book about it once. Canvas is good drawing API and it runs everywhere. However, despite the magic that is the Canvas API, it can still be tricky to use when it interacts with CSS. I often have people ask me how to make their canvas fill the screen, or resize with the window, or to have a fixed aspect ratio but still scale to fit inside the window without overlap. All of these require understanding some internal details about how Canvas works. So let’s dive in.

I’ve been teaching my nephew to code for the past few years and we did a game jam together over the winter break. He said to me: “Uncle Josh, you sure can code fast.” I reminded him that I’ve been programming for 30+ years, and one day he’ll be even better than I am.

In the last couple of years I’ve seen a lot of lamenting about the browser mono-culture. I even wrote about it myself. Some complains focus on how complicated the web specs have become. So big that only a few companies can implement a browser from scratch. I think these complaints are misplaced. Even if the web platform didn’t have such a large API surface it still wouldn’t matter. You can’t build a large scale browser with large marketshare. The browser market would still be a monoculture. You can't solve a business problem with a technology solution. I also don't think that replacing the web with something smaller like Gemini is the answer.

I've been working for a while on some new ideas around the future of programming, but I haven't done a great job of sharing these with the world. It's not science if you don't publish your results, so here we go.

How can you draw a filled rectangle fast? By making it not slow! Great if you are using a GPU to accelerate rectangle drawing for you, but what if you are doing it oldskool with an in memory frame-buffer? You'd probably write some code like this: