Notices by Adrian Cochrane (alcinnz@floss.social), page 35

Always aim high - Andy Bell: https://andy-bell.co.uk/always-aim-high/

Reverse-engineering the electronics in the Globus analog navigational computer - Ken Shirriff: http://www.righto.com/2023/03/reverse-engineering-electronics-in.html

@steeph I don't recall that conversation. Yes, that is basically what I'm building! Some nuance required, since there's some CSS I'm specifically deciding not to support since I think they cause more harm than good on the modern web.

Yes, it would run on desktop. But the UX isn't designed for that yet. If I can this working I might work on a variant where it is!

@steeph It's what I name Haphaestus's (TV browser) layout engine. It computes where to draw boxes onscreen, & this being the internet some of those boxes will inevitably contain cats.

Finally had a moment to read this statement from the FTC and it is 🔥🔥🔥

https://www.ftc.gov/business-guidance/blog/2023/02/keep-your-ai-claims-check

A few choice quotes:

>>

#AIHype #MathyMath #FTC

CSS4 Grid properties are now being parsed by CatTrap!

I think tomorrow I'll do some postprocessing converting the CSS data into the CatTrap datastructures. Since there's plenty of constructs which needs that...

After which it'll blogging & pulling everything together for a demo...

@passthejoe Shame it doesn't implement graceful degradation, I've heard some chatter appearing to suggest that makes it a harder to target.

I think they should a couple lines of code to avoid what webdevs take an interest in them from complaining about their limited CSS features...

CatTrap now parses the CSS display property, & I've prepared to parse the CSS4 Grid properties. I've created a place to parse to!

#ArgonautStack

@ebassi I do admit phone UX has led me to being overly-cynical about sandboxing in the past. I think I do have gripes against FlatHub, but those gripes applies to so much more than just FlatHub!

But as I implement my own browserengine I have to concur with your sentiment. Before WHATWG turned the web into an application platform, we didn't have great text rendering... Not long before that it was very euro-centric...

Plus, since it's an *early onset* form of brain rot, the "golden age" usually refers to stuff happening less than 10-15 years ago, which is ridiculous, and it's just a form of elitist gatekeeping in the most charitable of cases.

There's various types of (approximate) physics sims which frequently goes into computer animations!

Simulating fluids is fairly common, which even works well for fire.

"Particle" simulations are very common, with front-facing "sprites" being attached to each particle.

Kinetics simulations are usually used in videogames & (if your model is detailed enough) destruction scenes.

4/4 Fin for today! Tomorrow: raytracing!

If you want to render shadows, portals, mirrors, refraction, etc, etc using the depth buffering algorithm you must render the scene multiple times. Or incorporate raytracing in part or whole.

This is where it gets difficult to say whether depth-buffering or raytracing is more computationally efficient, it depends on what you want to do!

Shadows are accomplished by rendering from each light's position, then consulting depth-buffering in final render.

3/4?

In the early days 3D computer animation took cues from stopmotion & gave their 3D models skeletons. The bones in these skeletons could readily be converted into matrices to apply to selected vertices!

But if you want to actually render the otherwise-internal anatomy for characters like The Mummy or Two Face, have comically-muscular characters like Mr Incredible, or simply to animate faces you need to actually simulate the muscles. This made its way back to stop-motion via 3D printing.

2/4?

Hi, I work with generative machines. Everything from Markov chain generators to GPT-3. I’ve trained and tuned many models with GPT2 and 3, all with the intent of simulating human interaction.

I know a fair bit about generative machines, both how they work, and how to tune and interact with them to get particular results.

I need you to hear this: they do not know or understand anything. They are complex probability tables.

1/

To make depth-buffered 3D graphics more managable & impressive, there's several techniques commonly used with it. Today I'll cover several of these, without counting data entry techniques like Motion Capture.

Skeletal & musculature simulations aids animation. Multipass rendering overcomes some limitations of the concurrency strategy. Not to mention various physics sims.

Though the most important tool is keyframe animation, mentioned the other day as "interpolation"!

1/4?

There's abstractions around sort-parsing. And more abstractions around query parsing.

Introspective setter-callers.

Getters for the query's "purpose".

And seamingly nothing actually related to plugins, regardless of what this class's name says.

3.5/3.5 Fin! I'm not sure this thread's actually worth publishing?

When I next get back to this I'll start skimming over the entirity of Solr & seeing if there's anything worth commenting upon!

There's a utility for running the Solr query & command parsers, handing those off to the request's searcher, with error handling.

There's fields parsers.

There's a couple counting routines setting minimum thresholds, & wrappers thereof.

There's an optimizer for boolean queries. There's a string escaper.

Various regexes & wrappers thereof.

Quote balancing. Null-stripping. Datatype conversion.

A disjunction query-parser class, tracking a hashmap of aliases.

3/4!

The Fallacy of AI Functionality
“Deployed AI systems often do not work. They can be constructed haphazardly, deployed indiscriminately, and promoted deceptively. However, despite this reality, scholars, the press, & policymakers pay too little attention to functionality. This leads to technical & policy solutions focused on “ethical” or value-aligned deployments, often skipping over the prior question of whether a given system functions, or provides any benefits at all.” https://dl.acm.org/doi/fullHtml/10.1145/3531146.3533158

This plugin system includes a static mapping & a logger for global data.

It can fill defaults into network requests, with the aid of a support routine.

There's a utility abstracting results counting, & another to split words.

There's a routine for loading documents, unless lazy-field-loading is enabled, with or without calling into a match highlighter into the request's index.

There's a couple routines for managing debug flags. And routines to serialize debugging.

2/3?

The next component Linux From Scratch has you install into your new system is BinUtils, but since that's a vital part of the build system leading me to having already having covered it's complexities ( https://adrian.geek.nz/gnu_docs/c-compilation#gnu-binutils ) I'll instead study more of Apache Solr.

Apache Solr largely consists of a plugin system, & today I'll go over a utility class for managing those plugins.

1/?