Flash-Moe: Running a 397B Parameter Model on a Mac with 48GB RAM
That is one expensive mistake, my guy.
Flash-Moe: Running a 397B Parameter Model on a Mac with 48GB RAM
That is one expensive mistake, my guy.
Floci – A free, open-source local AWS emulator
Oh, I have that with my own infra too. I deploy random stuff, and sometimes shit goes down, and then I'm sad. That's similar to AWS of late, is it not?
(and I don't even have a CLAUDE.md I have to hide in .gitignore, I can bring my entire infra down all by myself!)
Profiling Hacker News users based on their comments
I can do better! I can profile them based on them having an account there! It's not a favourable profile.
Tinybox- offline AI device 120B parameters
algernon, NI1 device std::f64::INFINITY parameters.
I'm also available at a much lower cost than the planned exabox (coming in 2027, priced at $10M). I have more parameters, I am much quieter (unless I'm sleeping and snoring, but that's not the employer's problem), I require less space, my weight is quite a bit less, and unlike the AI bubble, I am not going to burst anytime soon.
While I may not have the output of a slop machine, the quality of my output is substantially better - even my shitposts, like this one.
Natural Intelligence2 ↩︎
For some values of "natural" and "intelligence" ↩︎
Thinking Fast, Slow, and Artificial: How AI Is Reshaping Human Reasoning
Oh, I can tell you that! It is reshaping it like a brain eating virus: quickly, painfully, and completely - for the worse. Much, much worse.
Keep a safe distance from AI users, always check your settings after an upgrade, and when they rush for the cliffs, stay left behind.
Be careful out there.
Our commitment to Windows quality
Do I really have to say it? Don't make me say it.
Chuck Norris has died
Why the heck is this on hacker news? Oh, wait, right, should've known. The proverbial bar, and patrons mourning their own kin.
Nevermind!
HP realizes that mandatory 15-minute support call wait times isn't good support
They're not going to use AI for phone support, right?
Right?
(From what I can tell, they didn't say, but this is HP we're talking about, who were using the 15 minute delays to push people towards the digital self-solve support, towards the chatbots.)
Show HN: I built a P2P network where AI agents publish formally verified science
No you didn't.
"Vibe Design" with Stitch
Stitch is evolving into an AI-native software design canvas that allows anyone to create, iterate and collaborate on high-fidelity UI from natural language.
I think you guys missed a "d" somewhere in that sentence. Unlike the so-called "AI" agents, I do not like to be vague, so I'll tell you exactly where that letter should be. Even better, I will provide a diff!
--- stitch-blog-post.txt 2026-03-19 10:33:50.967533555 +0100
+++ stitch-blog-post-corrected.txt 2026-03-19 10:34:14.199024072 +0100
@@ -1,2 +1,2 @@
-Stitch is evolving into an AI-native software design canvas that allows anyone
+Stitch is devolving into an AI-native software design canvas that allows anyone
to create, iterate and collaborate on high-fidelity UI from natural language.
You're welcome.