evilhackerdude

~a year ago i passed a major inflection point regarding web UIs.

i built an in-browser STT app UI with dear imgui, and everything just _worked_. i’m rendering to a canvas, running the model in a service worker and heavily using web audio, among other platform APIs. but.

at this point, i’d even struggle to comprehend the concept of going back to using the DOM for anything but a static reading experience. perhaps peppered with _fancy_ web components for an interactive animation. some would argue that’s what it’s for.

well, browsers/JS engines can both be powerful and terrible at the same time. despite using imgui via typescript bindings and rendering to a canvas via webgpu/webgl2, there is almost no limit to what i can do. more importantly, how easily i can express _whatever i want to see_ without paying with terrible frame rates, gc or layout jank or some other fun webism.

i’ve built SPAs since dojo was a thing, built iAds (lol), really, i’ve bikeshedded myself to oblivion.

low key expecting some kind of communal inflection point where we stop wasting our time arguing about bronze age tools and start having fun again.

i always thought alphabet’s own youtube videos must be a comparatively good source of new training data. if slop and other garbage is reliably filtered out it should leave plenty of higher quality content.

nothing like legendary shrugging and keeping the mind open

> The [next.js] developer experience is top-notch.

let me add my own unqualified statement to that: no.

> Next.js has invested heavily in Turbopack but if you want to deploy it to Cloudflare, Netlify, or AWS Lambda, you have to take that build output and reshape it into something the target platform can actually run.

it's almost as if vercel had some kind of financial incentive to gear this towards their own platform.

> reimplemented the Next.js API surface on Vite directly

a clown car screeches to a halt; several burnt-out-bored oracle vs google lawyers climb out and, weirdly, i am there for it

all in all, it's definitely a good example of something we couldn't have done for $1100 pre-llms, but: should we have? did somebody consult the lava lamps?

keep in mind that people who point out a regression and measure the actual #tok, which costs $money, aren't just "being loud" — someone diffed session context usaage and found 4.6 burning >7x the amount of context on a task that 4.5 did in under 2 MB⁣.

check out raylib, it should be amazing for this :)

sounds reasonable to me. i've been wondering about encoding detailed AI disclosure in an SBOM.

on a related note: i wish we could agree on rebranding the current LLM-driven never-gonna-AGI generation of "AI" to something else… now i'm thinking of when i read the in-game lore definition for VI (Virtual Intelligence) back when i played mass effect 1 ;)

i fixed a busted analog stick by opening it up and cleaning the sensor thanks to some youtube video.

i thought i could enjoy woodworking, but i was wrong.

i attended two 1-day woodworking courses: hand tools and machine tools. i made a nice bench/stool thingy that i still enjoy daily.

however, every time i work with wood, my mind immediately goes to CAD, 3d printing and wishing i had a CNC. i look at my soft baby hands (i can’t go bouldering for shit) and walk to the keyboard, create a parametric model in a notebook (build123d is great!), slice it and send it off to one of my printers…

sigh

I'm looking forward to the first native optimized WebGPU implementation of 3DGS rendering. I'm also curious how scene data could be compressed and decompressed efficiently.