Quinten Kock retweeted
If you are over the age of 25 and haven't yet been in a relationship, you will never be able to learn and master such complicated heterosexual concepts as The Toilet Seat Conflict. The neuroplasticity of your brain has decreased. Gay dating remains an option, however.
Quinten Kock retweeted
“If they wanted to, they would” is neurotypical propaganda and I’m tired of people pretending otherwise
Quinten Kock retweeted
wow branch mispredictions sure are slow and that’s such an obvious optimization, some enterprising compiler developer should surely ensure these compile to the same instructions (maybe even to something strictly faster than either!)
oh…
your CPU hates if. a mispredicted branch costs 15-20 cycles. every time. there's a version of abs() with 0 branches that most devs have never seen
x >> 31 arithmetic-shifts the sign bit across all 32 bits. negative numbers become 0xFFFFFFFF (-1). positive become 0x00000000 (0)
Readers added context they thought people might want to know
Virtually all modern compilers will optimise away the branch in abs_branchy, resulting in code that is at least as fast as abs_branchless. godbolt.org/z/WhvdnvaKz
Plus abs_branchless relies on implementation defined behaviour (right shift of signed integer).
>"One homepage can't serve them all."
>look inside
>one homepage
notice something?
Linear, PostHog, Attio - all shipped the same thing in the last few weeks. Homepage is a chat bar - not a dashboard.
This is the SaaS industry quietly admitting that traditional UI doesn't work anymore. Every user is different. One homepage can't serve them all.
The playbook is shifting:
→ expose your core APIs
→ connect an agentic layer
→ let users use software the way they want
SaaS became chat. Chat will become Generative UI - the agent won't just reply in text, it will compose the interface itself.
We're closer than people think.
77
Quinten Kock retweeted
Taking rest from functional programming by playing a horror game with global state, assignment statements and undefined behaviour.
Protesting against Trump is "actively stealing from their grandchildren", while Harris would have won if only people under 40 voted?
There's a lot of things you can say either way but I don't think "protesting Trump supports gerontocracy" is a coherent viewpoint.
Replying to @kenwheeler
Actively stealing from their grandchildren. Imagine being fucking 75 and going out into the streets with other old people to force your Summer of Love bullshit on everyone one last time before you die. It’s actually insanely sinister
51
Quinten Kock retweeted
Bloated patches: LM generated solutions of SWE-bench tasks are consistently longer than human-written gold solutions (and it's not just comments) 🧵
Kinda funny how the comments here are saying it's an undersold advantage; it's to my knowledge pretty much the entire reason @mozilla productionized Rust! Kinda makes me wonder what went wrong there image-wise; maybe "fearless concurrency" transcended from "benefit" to "meme"?
Nice thing about using Rust in @ladybirdbrowser: it becomes significantly easier to parallelize things when you have confidence that they're not racy!
Here's what it looks like when we distribute JS parsing to a thread pool instead of parsing synchronously on the main thread:
Genuinely hilarious how homelabbing has been a thing for ages, yet AI bros with their claw slop act like it's a new thing. Meanwhile their favorite tech companies are making sure that having one is crazy expensive by driving hardware prices up more than the crypto bubble ever did
67
Quinten Kock retweeted
Vim 9.2 is our first text editor that was instrumental in creating itself. The Vim team used early versions to debug its own .vimrc, manage plugins, and diagnose test results and evaluations—our team was blown away by how much Vim was able to accelerate its own development.
Quinten Kock retweeted
tbh this feels far more like brute-forced hill-climbing than the kind of insightful intelligence breakthroughs chollet hoped for this benchmark to incentivize
Replying to @arcprize
Johan's submission does a multi-model ensemble. It runs the same task through GPT-5.2, Gemini-3, and Claude Opus 4.5 in parallel. Tries multiple times with different prompting strategies (standard, deep thinking, with images).
Then, instead of predicting the grid directly, the LLMs write Python functions that describe the transformation rule, then execute that code in a sandbox to produce the answer.
After collecting many candidate answers, separate AI "judge" models evaluate and vote on which solution is most likely correct.
See the repo here: github.com/beetree/ARC-AGI
Dit moet zo hard gaan als je debiel bent
Replying to @WhitePapersPol
Going to add this early:
The Dutch could turn this around with a sensible remigration policy.
The country could be 85% Dutch, 95% European, and 5% non-European, mostly people connected to Dutch families.
125
Twitter is so funny because it's just full of right wingers constantly exposing themselves as brainlets
61
This is so pathetic lmao. I personally *do* think AI is cool to some degree on a base level, even useful in some cases, but requiring everyone to agree with or even affirm it has such a strong smell of insecurity. Forced "coolness acknowledgements" would be incredibly dumb.
I really don’t get why so many people can’t just even for a second appreciate how cool it is that we have sci fi intelligence simulators running around all over the place, even if they think it’s a net bad or they’re not very smart or this or that. Just take a second and appreciate how cool that is and then get back to your regular raging ok
2
4
5,337





































