Germán Molina Larrain

@LowLevelTweets But how do I get the security patches ? And the new insecurity ones?

this as fuck

@benhylak Is “an abomination” the right answer?

a lot of you don't know what a git submodule is, and it shows.

@plainionist Saying “make no mistakes” or “don’t lie” is not more context… it is asking it not to be stupid

Hot take: Most LLM hallucinations are caused by unclear prompts and missing constraints 🤷‍♂️

@juliarturc That sounds like the right solution, actually. Those are just silly questions… I want training data, energy and parameters used for smart things 🙃 not sure how sure I am about the above

What if the frontier models still don't know how many r's are in strawberry and they were just given a count-chars skill that calls a bash script to actually count them? 🤔

@rfleury @CodeEdison How about mine? I have been shipping slop since before it was cool

@CodeEdison Not your account

If AI writes 80% of code, who even survives in tech?

@GeeksGamersCom That is cheating. That trilogy is done, it is not fan art, it was not a “parallel universe” experiment… it was just a bad series of movies and that is that. It is like bad musicians cannot remove their bad music.

RUMOR: Disney to Remove Star Wars Sequel Trilogy From Timeline to Resume Focus on Original Characters "If true, this would be one of the most dramatic franchise shifts in modern Hollywood history."

@dieworkwear They write like this. Not like that. Not in different way. Like this… and it is annoying

Why do people write tweets like this? Where every sentence gets a new line. Sometimes a line might have two sentences. Like this one. But generally speaking, every sentence has a new line, making a tweet look like a long block of text that no one reads. Worse still, such tweets are often repetitive and winding, hammering on the same point over and over again. The writing is often very bad.

@paulg I learned a lot about how to give feedback when my PhD supervisors commented "Unclear" on some sections... like, please tell me what is it that is unclear, or add some questions or something. Otherwise, I do not know what to fix or how to debate.

You never need to describe something as a "bad take." If it's false or irrelevant or omits something important, you can just say so. And if it's true, relevant, and complete, it's not a bad take.

@shadcn 💯I do this

I see. How many of you do this? I’d like to get a sense of how common this is.

@JEBistline Tax data looks the same… bunching

With the Boston Marathon today, a good time to re-up one of the greatest figures in sports data: the distribution of marathon finish times (n=9,789,093). The spike at 4:00 is not a coincidence.

One of @Office's flagship products, BTW... copilot does not come up. But it gives me two Close buttons!

@ollama @thinkverse @googlegemma Can you share the exact model?????

@thinkverse @googlegemma vroom vroom!

.@googlegemma 4 on MLX in @ollama is way faster than before. 🔥

@ThePrimeagen Mehmehing

@ThePrimeagen Is this something that can be purchased?? Or just limited to omacon?

The new Dell XPS Linux battery test and HOLY COW, the battery life is EXCELLENT. A little background: Intel + dell + omarchy team worked together to get everything working perfectly for the new XPS. I am honestly shocked that I am having a "mac" experience with linux as far as battery life goes. Here is my test timeline: Saturday: 99% - 10:07am computer fully charged and on, walking to gate, suspend mode in backpack 99% - 11:00am - compiling rust, running agents, MiMo running on youtube, writing this tweet 80% - 2:00pm - forced to shutdown due landing the plane :( No more work, almost done setting up my machine! Very excited. Vim is there and so is tmux and zsh, but not my wall paper :( And i want to try Aether 75% - 2:30 - 3pm - i watched the new moist critical videos on the guy who threaten to kill people via ring doorbell. woah that was weird. Sunday: 5PM 75% -> 62% - Sat in suspend mode in my back pack for ~14 hours. I wanted to see where I was at, will open back up in another ~14 hours. Monday: 7AM 55% -> 5% - Monday: 3 hours of work. Agents are coding, neovim motions flying, youtube playing MiMo. Even took a 40 minute discord team video call and the microphone worked FIRST TRY?? Did... Linux just get a computer where you don't have to worry about battery life? It honestly felt better than my way back in the day Mac Air experiences. ---- Things I did not like: * when the computer wakes up from suspend, its "chunky" for about ~30 seconds. * the touch pad is annoying. the right click seems like its ~95% of the touch pad and i have a bit of a trouble getting left click regularly. The new Dell is actually good. I am shocked right now. Omarchy also took 3 minutes to install and I was up and running in 5. The primary reason why I am using omarchy is because 2 reasons: 1. everyone on my team is using it, makes certain aspects of life easier when everyone is on the *almost* same distro 2. intel + dell are working with each other and omarchy has a seat at the table to make things happen. this means i am using the super latest hardware with it perfectly integrated. pretty awesome. Thanks Dell for sending me the computer for Omacon! I am genuinely stoked for the computer.

@tris_does_stuff @marmaduke091 If you give me an equation or a complicated arithmetic problem, I would not predict the answer like an LLM does…. I would use a proper algorithm and math

@germolinal @marmaduke091 You're telling me that every time you're asked "What's 2+2?" you do the actual math in your head? You don't just remember the answer?

>AI can't do math >Look inside >Strongest model they tested was GPT-4o-mini Get this slop out of here. The models tested are so inferior to current SOTA models that the findings are completely irrelevant to the current AI landscape.

@jordwalke 1. I have seen worse rust pattern matching syntaxes 2. My main use for enums is that the compiler can tell you if you are missing some branch (eg, you add a new error type and it nudges you to update)… although that will not work on your example because you are matching the _

@interesting_aIl Does this assume the robot vacuum goes in a straight line?

How long it would take to reach the Moon at different speeds

@rezoundous touch grass, hopefully

In March, people were flocking to Claude. In April, people are flocking to Codex. Where are we going in May?

@glcst Aaaaand, compilers/assemblers are deterministic... not like those temperamental fast-typing monkeys that LLMs are.

Lots of people compare LLMs to assembly, and say "You didn't look at the assembly output of your program, why would you look at the output of the LLM?". Well, first, in the Linux Kernel we did look at the generated assembly often. But leaving that aside: Except for the rare compile bugs, I can trust that if I write correct code, I will end up with correct assembly. I also know that the compiler will not start adding security holes on its own volition. No such guarantees with llm.