Just kidding. But you must have clicked somewhere. Audio autoplay hasn't been a thing on the web for years (thankfully).

For example, in the Breaka Club (https://breaka.club/) editor, which is not presently exposed to kids yet, we define behaviors for characters, props, mosaics (terrain tiles) and items in JSON. But the JSON has type checking and real-time completion as you type. Not naive property auto-suggestion, we have effect types and they're context aware in that you can refer to targets introduced higher in parent constructs of the JSON — and they're fully type checked. If you refer to a target that does not exist in a context, it won't validate and you cannot save the schema.

We use tsgo for development, but the editor (runtime schema validation) is presently stuck on TypeScript 6 because there's no WASM support.

Now, is it nutty that we're using TypeScript for JSON validation? A little. But it's extremely powerful. We're going far beyond what's capable with Zod or ArkType.

In particular, I'd expect there to be shaders making use of textures. I was expecting buttons with a laser finished metallic look that actually changes with lighting on the page. Or hero images that combine photography with masks to produce interesting effects highlighting products with auras etc. These all appear to be screensavers.

~/.codex/config.toml

  model = "gpt-6.1-sol"
  model_context_window = 700000
  model_auto_compact_token_limit = 630000

That said, I'm not able to get anything useful done in those hours. The exception is when I swap back to 5.6 Sol/Luna, which seem to be working as they used to.

Mostly, I've just given up and am using Opus 5.5 for all the important tasks now. Which is odd for me say, because I cancelled my Anthropic sub a few months back because I couldn't stand Opus 4.6 < x < 5.5. 5.5 feels great though, like Astra did when it first released.

I don't know what conversation you think you're having here, but it's not that.

If you'd bothered to read... and then ready for it... stop, take a breath, and process the information. You'd have noticed I was pointing out you can use a Quest headset WITHOUT providing them your personal ID OR having a Facebook account. Almost as if, you know, I was highlighting that giving the company your personal info isn't required. Woe me though for pointing that out.

I'm sorry for interrupting the indignant rants. Please continue.

https://breaka.club/blog/why-were-building-clubs-for-kids

Have been running an in-school pilot over the last school term. Has been super insightful and led to a heap of changes to optimize UX. However, the biggest challenge has been delivery in this environment.

Trying to run a game in a browser on school iPads powered by really locked down networks has been challenging. Some school iPads will crash (out of memory) when launching the camera view, whilst others run flawlessly. We use a fork of Godot, and are constantly fighting to strike a balance between UX (increased parallelism) and browser device constraints i.e. a limit on how many web workers we can fit in memory.

In addition to our main (build and play your own RPG) experience, we also stream games direct to iPad. In particular a modded version of Overcooked! 2 that teaches kids to code — Overcooked itself was never even released on iPad. We've built our own Kubernetes system that spins up (a license limited number of) games/pods on demand backed by (time sliced) NVidia GPUs. This is literally running out of my home office. Despite the complexity of building all that, the Kubernetes and GPU time slicing wasn't even the largest hurdle. Turns out UDP is a no-go on the school network, so direct WebRTC was out the window. Instead we're using Cloudflare's TURN relay with TCP. Which much to my surprise, is holding up quite well.

Definitely a learning experience. However, kids are having a blast. Can't wait to roll this out further!

https://www.youtube.com/watch?v=ITWSL5lTLig

The video above shows keyboard, but the main motivation for the key-based navigation was actually gamepad.

https://breaka.club/blog/why-were-building-clubs-for-kids

The primary mechanism through which kids create their game worlds is taking photos of their own hand drawn artwork.

I found traditional game dev tooling was far too technical and verbose for my daughter to stay interested. Honestly, it's not just a matter of her losing interest due to the complexity. Frankly, I don't want her sitting in front of a screen that long. She's a kid and I want her doing typical kid stuff. She and her younger sisters play sports, draw, paint, perform magic shows, and sing and dance around the house like complete nutbags — and I love it!

So whilst what I've built is highly technical for me to maintain, I'm not trying to expose all that complexity to kids. At least, not yet. I'm mostly interested in fueling creativity and letting kids explore. That's not to say kids that want to dive deep can't — we offer visual coding experiences etc. The intent is kids gain access to new features through self directed exploration and play. If some kids stop at just play, that's OK. The primary game mechanics (drawing and building worlds) are still inherently creative.

Although my approach is different, there is a heap in this article I resonant with. The hacking mentality, and letting kids go beyond the artificial constraints in particular. There's still much of Breaka Club that I can improve on. I'm running a limited in-school pilot at present and that's already yielded some really interesting insights. I'm really keen to continue to see what others put out in this space. I think it's drastically underserved and there's plenty of room for multiple approaches.

https://breaka.club/blog/why-were-building-clubs-for-kids

I also attempted to make a game with my eldest daughter initially. Unsurprisingly, she didn't want to sit there watching me code or drag things around in Godot's level editor.

Admittedly, I really need to make a new blog post showing the latest improvements brought about from the in school pilot I'm running weekly at the moment.

We stream OC2[1] with our mod preinstalled over WebRTC. This ensures that kids/schools don't have to try and install the mod. This is particularly important since we support running on school provided hardware. Installing a game without a mod would be hard enough. Added advantage though is kids play with a virtual (on screen) gamepad on iPads in Mobile Safari.

Game instances run in Docker containers in Kubernetes/k3s atop very outdated nVidia hardware. Given we're already going across the Internet into school networks, we've tried very hard to optimize latency across the board. Using NVidia NVEnc with DMABuf (zero copy) etc. We're unfortunately using XWayland at present so experience the documented input overhead. Although our inputs are virtual devices at this point, so the overhead may be a bit different. Trying to optimize this whole thing end to end has been a challenge. I would say that performance is currently "acceptable".

OC2 coding: https://www.youtube.com/watch?v=ITWSL5lTLig (not streamed in this case)

[1] We've bought a limited number of copies of OC2 and pods claim a license on startup. If we're at capacity, kids play something else.

However, as a senior engineer with fairly deep technical knowledge in these areas, I'm now using AI in my own projects. Frankly, before even touching Github, I'm already drowning in code reviews generated by my own use of an LLM!

There are meaningful, generally good PRs waiting for my attention against GodotJS, and other projects I (somewhat) maintain e.g. MoonSharp, C# Lua runtime. It was already extremely difficult to stay on top of PR code reviews for reasonably technical projects. When you're already somewhat burnt out from reviewing LLM code all day, it's so much more exhausting than it used to be.

https://breaka.club/blog/why-were-building-clubs-for-kids

We also teach kids visual scripting in Overcooked 2!, allowing kids to code their way through the levels of an existing much beloved game:

https://youtu.be/ITWSL5lTLig

I'm running an in school pilot this week (Lunch time school club).

The tech stack for the main product is honestly pretty intense at this point with full multiplayer support, offline play, transitioning from client authoritative to joining a remote server. Built atop GodotJS, TypeScript bindings for Godot, which I maintain. Huge monorepo with over a million lines (yes, I'm aware that's NOT a good thing), and GodotJS itself is not included in that.

If I'm at the point of contributing a PR to a dependency, I've already identified the root cause in detail. There's no way a change log should be going into that level of detail, or else you're just duplicating the Git log for no reason.

Will the change log make mention of fixing the bug? Perhaps. But I'm going to want to read the technical details of the fix to make sure they've specifically addressed my issue, and not just a similar problem. What is the performance impact of the fix? Are there security implications they've explained in the commit message.

I'm a software engineer, not an end user, I want the technical details of my dependencies.