Totally agreed, most python package management tools suck. But uv is actually incredibly good now. It’s insanely fast and just works without weird errors or footguns.
https://www.bitecode.dev/p/a-year-of-uv-pros-cons-and-should
I would like to publicly say I was wrong on this one. According to https://lmarena.ai, Grok 3 is the best for style at least. For content GPT-4o still edges it out but I was def wrong that xAI wouldn’t achieve SOTA.
Programmers: “A few billion tensor operations a second to simulate language? No problem.”
Also programmers: “A calculator app? That’s going to take a PhD in mathematics and some new research to almost get a correct calculator.”
Totally agreed that popularity doesn’t mean it’s good. I’m simply saying that veganism being a popular diet is the more likely reason that Nat Geo is writing an article about it instead of a psyop
With all the anti science stuff going on in the media, here’s your reminder that there has never been a competent study at realistic levels showing any downsides of fluoride. On the contrary, many studies have shown no health concerns and large dental health improvements from fluoridated water.
Stop believing whatever media personalities or podcasters tell you and do your own research. Remember, bitcoin is built on verification, let’s follow that example.
https://www.scientificamerican.com/article/fluoride-in-drinking-water-is-safe-heres-the-evidence/
Veganism is a relatively popular diet. I agree with you that veganism for dietary reasons is not the best option, but I don’t understand how this is a psyop.
That’s an interesting theory. Google is waiting to trounce OpenAI until the Justice Department calms down? It seems like Google should be able to annihilate OpenAI, since they have access to a larger user base, more products to integrate, and an existing cloud infrastructure and ecosystem that devs already use.
Google keeps releasing new and better models and AI tools before OpenAI at lower prices and yet no one even knows. Google’s AI division makes great products, but they’re awful at announcing them and getting publicity. They need to have at least some sort of big announcement and fanfare. Get up on a stage, make a keynote, a demo, literally anything.
Example: deep research. Google released the same product with the same name, but two months before OpenAI. Everyone still talked about OpenAI’s announcement and acted like it was totally new.
https://blog.google/products/gemini/google-gemini-deep-research/
Found this yesterday. https://marketplace.visualstudio.com/items?itemName=copilot-arena.copilot-arena
It's the VSCode extension that powers the Copilot leader board on LMArena. I looked through the source code. Seems pretty simple.
I think I'll take a crack at porting it over to Neovim.
After some thought, all new projects will use Typst and I can just switch old projects to just use Typst as the pdf engine instead of LaTeX because the Typst compiler is stupidly fast. One small CLI flag change in a Makefile and compilation is now stupidly fast.
nostr:nevent1qqs2yjuug0gh4lsk0phww0jayeh00kcp54fzk4y4egac9f7j2hcyp2gpp4mhxue69uhkummn9ekx7mqeyk3ek
Considering moving from a hacked together setup of Pandoc Markdown, Makefiles, Python scripts, and a sprinkling of Jinja templates and raw Latex to Typst.
Does anyone have experience switching from Markdown to Typst? To me it seems like it optimizes well for making the hot path of the markup used for most documents look similar to Markdown while also allowing you to drop down into lower levels of abstraction seamlessly.
Any footguns or horror stories I should know before I put in the effort to switch?
I know, I always find the entire Buy Now Pay Later market to be shockingly large. I suppose it makes sense that as time preference goes up products that let you get stuff faster will become more popular. Still, I’m surprised how many people thoughtlessly take on debt for consumer items.
Maybe, but remember that social media will give you a warped view. You’re generally only interacting with English speakers, which will bias your feed towards Americans but also bias you towards assuming they’re Americans.
The equities and ETFs are a solid argument. The US has nearly half the equity market globally. Computers were also widespread here in the late 2000s so we prob got solid early adoption.
Maturity is realizing that while Elon does actually deliver incredible products and innovations, he still manages to overshadow the real advancements with the sheer amount of claimed announcements that never happen.
The principle honestly goes for most Musk statements. 19 out of 20 times he’s spewing total bullshit but 5% of the time they do something crazy like catch a rocket. 5% is a solid base rate for any Musk statement actually happening.
Forecasting tip, whenever Elon says something about xAI, immediately bet that it won’t happen. Literally every time they release a model everyone on twitter huddles up to hype it and then it comes out as the most bland model that was already outdated before the training run began.
Follow for more investing tips lol.
Maybe, but the S1 paper showed that to really achieve top level performance you really only need an extremely small dataset of high quality reasoning examples. That’s enough for a normal LLM to pick up on the patterns and you’re good to go.
Well, not much of a game.
