#Autocomplete
“An Autocomplete Psychotherapist with no duty of care”
November 11, 2025 at 1:10 AM
thinking about standing up a base model raw autocomplete llm server but the options are "so small it's really dumb" and "this will cost real money" and that's a lot to commit to something i am doing for grins
November 11, 2025 at 10:30 PM
monsters from the id
November 10, 2025 at 1:47 PM
the issue is not that the fancy autocomplete will be good enough to replace workers, its execs thinking they can keep their company running with probability machines while saving money on salaries
November 12, 2025 at 6:52 PM
A clip from the autocomplete interview with three very important staples of my personal stanhood.
1. Being extremely Taika-ish here
2. SHOWCASING one of my favorite features in his accent
(The diphthong in Know/Show - knoieuhw 🥹)
3. Gratuitous Hand Thing ✊
November 12, 2025 at 6:48 AM
Typed "male" into Amazon and the autocomplete is fun
November 9, 2025 at 6:06 PM
Reason. No particular reason. I swear the word was there. My brain has autocomplete.
November 12, 2025 at 1:35 AM
November 10, 2025 at 1:26 AM
I'm really excited to release the next iteration of the #WebComponents language server! This comes with some solid improvements:

- Autocomplete for #css variables, parts, and states
- Basic binding integration for popular frameworks, including Angular, Vue, and Lit.
- Lots of perf improvements
November 12, 2025 at 1:00 PM
no i mean this would be like. if i went and got the deepseek base and just left an open autocomplete server to demonstrate how weird base models are. the problem is this would cost like a hundred bucks a day if i did it cheap
November 11, 2025 at 10:35 PM
It’s hard to take something seriously when one of your coworkers says ‘we definitely need to do what an autocomplete trained on Reddit says we should’ whenever you bring it up.
November 8, 2025 at 2:14 AM
Friends of mine who know much more about this than I do call it spicy autocomplete, and I trust their judgment.
November 8, 2025 at 12:55 AM
This is like trying to argue with text autocomplete on why it keeps substituting "ducking" for "fucking".
if i ever reach this point then just put me down
November 7, 2025 at 2:51 PM
How? How on earth could AI “end scarcity”? What people are calling AI now is fancy autocomplete. It doesn’t produce anything. It’s not a fucking replicator. What the hell are you on about?
AI could end scarcity, end humanity - or boost trend growth by 0.2 percentage points
November 7, 2025 at 2:46 PM
Me and @czwolf.bsky.bark.wolp.chat have known each other so intimately for so long that we can straight up autocomplete each other's sentences mid conversation, despite us being completely in different tasks at different locations
November 12, 2025 at 7:04 PM
Curious for folks who have gone from an M1 series Mac to an M… higher: does Xcode's AI autocomplete load any quicker on the newer chips? I yearn for Cursor level speed but I assume Xcode is trying to do it all locally?
November 7, 2025 at 7:01 PM
Strands #619
“How puzzling”
🟡🔵🔵🔵
🔵🔵🔵🔵
Very meta in an industry way. @jepaco.bsky.social @geoffbrock.bsky.social My handle autocomplete has returned!
November 12, 2025 at 9:21 AM
goddammit you guys finally have an idea that wasnt fed to you by The Spicy Autocomplete and its just blaming the jews again? its been 100 years, find new material!!!
There’s no daylight between what Candace says here about “the Jews” & what Nazi propagandists & Father Coughlin said about the Jews in the 1930s. Her show is one of the top 5 “conservative” podcasts in the US. She has 5.5 million subscribers on YouTube and 7.3 million followers on X.
November 8, 2025 at 9:29 PM
i don't mind using e.g. LLM autocomplete because i know what i'm doing already and it's better than the autocomplete i had before. but i also know ppl who just ask claude to make a feature and make a PR and it's 1000 lines of spaghetti that they never even looked at
November 8, 2025 at 2:50 AM
They don't know how to do that because they don't know how to modify the parameter weights to achieve anything specific. All they can do is reinforcement learning. There's no chat algorithm for anyone to tweak. It's fancy autocomplete and it has no concept of truth, knowledge, or what it's doing.
November 7, 2025 at 3:56 AM
I know at $previousjob we were measuring autocomplete suggestions accepted which made me angery
November 6, 2025 at 5:56 AM
In a world where there wasn't the visceral backlash (and they were resource-lighter, which I guess would help), it would be hilarious to have constantly-running autocomplete that, e.g. completed your thought but always in pirate speak, or worked out a segue to The Fresh Prince of Bel-Air, or...
November 11, 2025 at 10:40 PM
This is a great example where Cross-Root ARIA would be great for progressively enhancing an input into a combobox with autocomplete.

fwiw, I hear `<input><datalist>` is getting an `appearance: base` overhaul and you could potentially wait for that to save lotsa code... but same x-root ARIA issue.
November 11, 2025 at 4:11 PM
Yeah, nothing like getting confidently lied to by an autocomplete lol
November 7, 2025 at 12:21 PM
Went to type 'holy shit' and my phone tried to autocomplete Holy as Goku
November 7, 2025 at 6:26 AM