Notices by Adrian Cochrane (alcinnz@floss.social), page 63

Many of most recent links have come from...

Book production, AI, Single-Page-Apps, and other links and notes - Baldur Bjarnason: https://www.baldurbjarnason.com/2023/book-production-ai-spa/

Some good additional commentary in there!

The User Activation API - WebKit: https://webkit.org/blog/13862/the-user-activation-api/

This relates to my approach to gaining meaningful consent, my main difference is I seek it for more fundamental things like communicating over the network. Once I've dropped all these JavaScript APIs...

Uncontainable Container Query Excitement - Matthias Ott: https://matthiasott.com/notes/uncontainable-container-query-excitement

It's not an immediate priority for me, but hearing the excitement & judging the added complexity of this feature to be as low as possible given its purpose I do plan to implement it!

Dev Journey #240 Chris "Go Make Things" Ferdinandi from HR professional to vanillajs-guy - Timothée (Tim) Bourguignon: https://devjourney.info/Guests/240-ChrisFerdinandi (podcast episode)

ChatGPT Is a Blurry JPEG of the Web - Ted Chiang @ The New Yorker: https://www.newyorker.com/tech/annals-of-technology/chatgpt-is-a-blurry-jpeg-of-the-web

Automation Isn’t Automatic - John Warner @ Inside Higher Ed: https://www.insidehighered.com/blogs/just-visiting/automation-isn%E2%80%99t-automatic

Magazine Publishes Serious Errors in First AI-Generated Health Article - Jon Christian @ Neoscope: https://futurism.com/neoscope/magazine-mens-journal-errors-ai-health-article

3 thoughts on a decade of publishing books - Austin Kleon: https://austinkleon.com/2018/05/22/3-thoughts-on-a-decade-of-publishing-books/

Human curation - Manuel Moreale: https://manuelmoreale.com/@/page/MjoDjqUzICR1ne4h

Hear, hear!

Screen Readers Don’t Convey the Semantics of STRONG and EM - Matthias Ott: https://matthiasott.com/notes/screen-readers-semantics-strong-em

Rhapsode on the other hand...

Tends to sound especially robotic on italicized text, so definitely keen to hear advances in voice synthesis I can use! Funny that the AI more human-sounding voices don't give me these knobs...

In much of the advocacy for "plain English" writing, the only real disagreement I have is the opposition to jargon.

I get the impression this is mostly coming from managers who dislike their own jargon & probably have good reason for it. In my & many other fields, given the right audience, jargon helps a heap with clarity & concision! Isn't that the point of Plain English?

Well, when writing hypertext I can link to definitions...

Don't know why this came back to my mind today.

this isn't really an insight, but more of a revelation of framing things

when people say we no longer need (artists, writers, "to write academic papers", etc....) because chatGPT will do it for us

those people almost always *self-describe* as being uninterested in either *producing* or *consuming* those things

i guess i'm just stuck on the implications of an AI scholar with avowed antipathy toward the material artifacts of our work - the conference papers, journal articles, blog posts, etc...

every time I start a new frontend JS project I consider using a bundler (like rollup/esbuild/etc), and then I think about how happy I am that my past self didn't use a bundler because my old projects are so easy to get working

Reading FreeType documentation, it seems I really should have read these pages before I reimplemented Typograffiti text rendering! It warns me about issues I'm having as I test just a tiny bit more!

https://freetype.org/freetype2/docs/glyphs/glyphs-2.html

O.K., get font sizes & baselines working tomorrow... I've learnt the necessary jargon to do so: ppem, ascent, & descent!

New release of Typograffiti coming soon, which should look much more legible!

#ArgonautStack

my verdict on this after fucking around with it for about a day: it's just really not ready. there's a huge gulf between "can impressively do some stuff" and "can actually be used at all in any capacity" and it's firmly on the useless side currently. i've seen a lot of discussion about how it "sometimes" hallucinates stuff or "occasionally" makes mistakes that catch you out. but it isn't occasional, it's constant! e.g. every non-trivial bit of code tries to use an API call that doesn't exist etc

@EcoActivist @eevee @jk Large Language Model. The name basically translates to "it's like autocomplete, but on a much bigger training set"!

More search results for your enjoyment: "Female portrait."

God, I wish I was fucking kidding.

Ugh, found another bug to fix in Typograffiti! Turns out the scaling factor on the units Harfbuzz gives me is dynamic... I'll be doing some reading on this, since I didn't what I was doing before...

Probably investigate my baseline issue while I'm at it...

#ArgonautStack

I've just drafted a new CatTrap API, and have just committed it. Now I shouldn't get stuck on implementing this one...

https://git.argonaut-constellation.org/~alcinnz/cattrap

This'll definitely be worth a blogpost when I implement it, give webdevs a clearer picture of what these CSS properties do!

@RL_Dane @prometheus Yeah, reminds me of the aversion to tactile buttons many modern devices seem to have... Touchscreens are cool aparantly!

The trickiest units to support may surprisingly be percentages! Since that has a cyclic data dependency to break.

During initially computing min size & padding the spec says I should treat the parent as being 0 wide. max & preferred size would be based on the natural width.

Oh, for autolayout I'd need a postorder traversal to sum fixed/finalized heights.

---

With a minor change to underlying datastructures, I could write accessors for computing min, max, natural, & final width & height!

2/2

Let me talk through my CatTrap layout engine?

The minimum & natural sizes can be computed in a postorder traversal. A preorder can compute the maximum widths.

Once I have the min & max widths I can clamp the natural widths to them.

Once I know the final width I can compute the natural height in a postorder traversal. A preorder traversal can then finalize it within the given bounds.

That'll give good input to more advanced layout. And yes, I'm not tackling vertical text yet...

1/2