dorian

yeah i always thought it was funny how balsamiq had like fake sharpie marks and comic sans but also had smart guides and stuff, so you could arrange the intentionally janky widgets with mathematical precision

yeah balsamiq etc; dunno if buxton worked on one that went to production or if he just advised

heh i'm in a not-too-dissimilar situation with sense atlas and intertwingler, and like "making a product for other people" is an entire separate project on top of "getting the damn thing to work at all"

(i was not deluded about this part but maybe how much effort it would take)

isn't it so excellent?

i just have a ream of cheap copier paper from staples on my desk and a box of bic mechanical pencils lol

oh yeah ps he published that book 20 years ago

that would be 👨‍🍳💋🤌 and fully justified by whatever means they used

ah the 'I don't like X >> it's because X is a and b >> a and b makes X morally wrong >> no one should like or do X >> those who do X are both stupid and evil' pipeline

most confusing part is that it genuinely do be like that sometimes. not always though. not even most of the time.

bill buxton wrote an entire book to address the problem that on a computer, less polished looks more polished, and so stakeholders will assume that what you're showing them is much farther along than it actually is so you have to spend effort making it look sketchy

www.amazon.com/dp/0123740371

the smoothest brain

Man Kampf

Artificial intelligence, a marketing term that essentially refers to a computer that does the right thing in its situation without explicit programming. AI can be either deterministic, or statistical, or both. Machine learning, a family of techniques that use statistics derived from training data, which got popular over the last couple decades due to the internet and concomitant abundance of said data. Artificial neural networks, a particular strategy of machine learning that can turn a fuzzy input into a sharp output, at the risk of occasionally picking the wrong output. Language models, which are just neural nets that primarily operate over text. Large language models, which are just really expensive variants of the preceding, trained on data which was unscrupulously sourced. Chatbot-as-a-service, a specific category of commercial product furnished by the likes of OpenAI, Anthropic, and Google.

(i was on the hoof but now back at my desk; it's this)

Artificial intelligence, a marketing term that essentially refers to a computer that does the right thing in its situation without explicit programming. AI can be either deterministic, or statistical, or both. Machine learning, a family of techniques that use statistics derived from training data, which got popular over the last couple decades due to the internet and concomitant abundance of said data. Artificial neural networks, a particular strategy of machine learning that can turn a fuzzy input into a sharp output, at the risk of occasionally picking the wrong output. Language models, which are just neural nets that primarily operate over text. Large language models, which are just really expensive variants of the preceding, trained on data which was unscrupulously sourced. Chatbot-as-a-service, a specific category of commercial product furnished by the likes of OpenAI, Anthropic, and Google.

(i was on the hoof but now back at my desk; it's this)

i came up with a six-bullet stack to disambiguate it

2026-03-06 WORKING IN PUBLIC: the checkout continues YouTube video by dorian taylor

from this morning: going over Semantic REST, the lecture series, Sense Atlas, and using it to design the checkout flow: youtu.be/anMeXQlyjVM

(the voice channel i mean; assuming that's a general acl thing rather than me specifically)

it said i didn't have access to the channel btw

The Making of Making Sense So it basically goes like this: Assume my tweets skeets are shitposts. Assume my website is serious business. This newsletter is the interstitial fluid between the two.

in a forthcoming newsletter i coined the term P(DWIM), the probability of claude code doing what you mean, and the DWIMability field, which acts on the space of possible outcomes

(gonna try to get that out this weekend; you can subscribe here: buttondown.com/dorian)

and it appears that the sucking sound you're hearing is the vacuum of those higher-level decisions

i mean, sort of

the thing about that scene is that it's kinda contrived, like he could have stopped at the shelf and gone back to it later

aaand crash

well, hang

i think it might be time to put exwm out to pasture; it's just causing too much trouble

what seems to be happening is some kind of feedback loop that runs the system load up to like 283 and locks the entire thing out so i have to kill it from ssh on another machine

gonna do another planning session this morning on this checkout flow after working NINE HOURS last night to fix a bug in sense atlas so i can resume using it to do said planning

• stream.place/doriantaylor...
• twitch.tv/methodandstr...
• youtu.be/anMeXQlyjVM

gonna do another planning session this morning on this checkout flow after working NINE HOURS last night to fix a bug in sense atlas so i can resume using it to do said planning

• stream.place/doriantaylor...
• twitch.tv/methodandstr...
• youtu.be/anMeXQlyjVM

i swear "just gotta do this thing and then i can get back to what i was doing" but like recursive

make that nine hours jeeeeeesus

me at 3pm: i'll just fix some stuff quickly

wallace shawn should reprise his role

i am sure this is done to death but

"during his second, history-making, non-contiguous term in which he unambiguously won unlike the first time around"

love how i have been working on a pagination dereferencing bug for the last three(?) hours