bionhoward

Personally, I don’t think of it as fighting, more like “compiler assistance” —

you want to make some change, so you adjust a struct or a function signature, and then your IDE highlights all the places where changes are necessary with red squigglies.

Once you’re done playing whack-a-mole with the red squigglies, and tests pass, you know there’s no weird random crash hiding somewhere

Sounds like a good use of Num

[1] https://docs.rs/num-traits/latest/num_traits/trait.Num.html

The raw CommonCrawl has 100 trillion tokens, admittedly some duplicated. RedPajama has 30T deduplicated. That’s most of the way there, before including PDFs and Alibaba’s other data sources (Does Common Crawl include Chinese pages? Edit: Yes)

Wouldn’t it be 1GB (billion bytes) per billion parameters when each parameter is 1 byte (FP8)?

Seems like 4 bit quantized models would use 1/2 the number of billions of parameters in bytes, because each parameter is half a byte, right?

the whole point of AI companies right now is to suck up as much data as possible and figure out what the AI can do with that later.

If, later, that capability level reaches a point where they can take people’s jobs, then the data which noobs PAY to give to AI right now, will absolutely help them do exactly that.

And the paying noobs, agree to the terms, which say they aren’t allowed to use their own chat logs to compete with AI — literally the worst, most retarded deal in history, and millions of idiots think NOT taking it is dumb?

“If you’re not skillmaxxing with ClosedAI hours every day, ngmi” is the manipulation du jour for these assholes

Ironically, the header for Anthropic’s coding tool implies anyone who uses it consents to having Anthropic reverse engineer every ounce of code you send to them for every response.

How can a crew that steals all of everyone’s work and then says you can’t use their work for anything competitive with their “do anything machine” be considered anything except a bunch of unethical hypocrites?

It’s honestly sad how many people are paying to get pwned, dependent, and replaced by big AI companies right now

Couldn’t it wind up being easy for Cursor and other variants of VS Code to be long run beneficial for VS Code itself? Seems like having a different third party team extending your stuff and testing it, could be hugely valuable, they take risks and move fast, the upstream project gradually learns from what works for the forks, people contribute various other new extensions.

In the age of LLMs, community is worth its weight in platinum, cutting off Cursor just incentivizes them to develop some new better thing with better technology (cough Zed, Ghostty) to compete with VS Code which won’t benefit Microsoft because it’ll be separate. What’s the use in not just open sourcing the C extension? With more people moving off C anyway, might as well get the free community contributions

Seems genius, hope this catches on

step 1: don’t use the JVM

Praying to the kernel gods for some Rust FP8 training

LM Studio seems pretty good at making local models easier to use

sounds great as long as you don’t make any product or service that competes with Claude. Can anyone name something in that category?

“I built an AI”

Looks inside

REST API calls

To take your codebase and make you dependent

at least using llama is less dumb than using the Gemini, ClosedAI, Claude, or xAI apis, since it’s not effectively a one way wiretap like those lemming tier options

they just want to steal your whole codebase to train AI and then turn around and prohibit you from using it to code competing products or services. You know, the OpenAI/Gemini/Anthropic/xAI strategy

What a time to be alive!

your example shows using OpenAI, but they have a prohibition on competition, so what are we supposed to use in production?

I’m hopeful the Windsurf folks see reason and refuse this deal, because OpenAI’s prohibition on learning is absolutely insane to force on devs, isn’t it?

Claude and 4o aren’t RL trained IIRC? Also, who’s using these for code? You’re cool not being able to train on your chat logs used to develop your own codebase? Sounds pretty sus

Did the article mention the part about how these companies turn around and say you’re not allowed to use the output to develop competitive models? I couldn’t find mention of this

IMHO yes you’re “losing neurons” and the obvious answer is to stop using Claude. The work you do with them benefits them more than it benefits you. You’re paying them to have conversations with a chatbot which has stricter copyright than you do. That means you’re agreeing to pay to train their bot to replace you in the job market. Does that sound like a good idea in the long term? Anthropic is an actual brain rape system, just like OpenAI, Grok, and all the rest, they cannot be trusted

AI studio is only developer focused if you’re not working on AI, which is a prohibited use case according to the Gemini API / AI Studio “Additional Terms”

IMHO, aren’t modules necessary for big projects to limit the amount of complexity we have to deal with at any one time?

Our minds can (allegedly) only handle 7+/-2 concepts in working memory at once. Your whole codebase has way more than that, right? But one module could easily fit in that range.

HR wranglers? Damn, that’s simultaneously hilarious and terrifying

How does this compare with Byte Latent Transformer [1]? This happens with convolution post-embedding while BLT happens with attention at embedding time?

1. https://ai.meta.com/research/publications/byte-latent-transf...

Avoid it! Perplexity rules prohibit commercial use, so it’s effectively a toy for hobby stuff like gardening and recipes. ChatGPT rules prohibit sharing the output with other AI, and they’re training on everything you pass in, so you’re literally paying to get brain raped, they hold all the cards, and that’s as dumb as it sounds.

Anybody coding with OpenAI at this point ought to be publicly mocked for being a sheep. All you’re doing when you code with AI is becoming dependent on something where you’re paying them to train their model to imitate you. It’s not good for anyone but OpenAI.

Better off learning a better (compiled) programming language. If you have to use AI, use Groq or Ollama as you can keep these outputs to train or fine tune your own AI one day.

Why pay OpenAI for the privilege to help them take your job? Why outsource mental activity to a service with a customer noncompete?

Wrong about OpenAI because it has a customer noncompete

Better docs would be appreciated! Can you guide us to use it? I’m not a slack guy because it’s too distracting, but easy secure chat integrations could be big … keep it up!

Depends what you’re doing, but I think these are sensible defaults:

HTML/CSS/JS won’t go out of style (could use templates like Askama, utility classes like tailwind to speed things up)

Axum is a nice option for backend because it’s much less buggy than the Python/JS stuff

Redis / Valkey for a KV cache

PostgreSQL database

Anthropic didn’t build shit, it’s got the same closed output rules as ClosedAI and xAI and Perplexity and the Gemini API. Send them your precious questions and code and what do you get back? Output you with a prohibition era policy telling you not to use it to compete with their thing that does everything. That’s such a dumb deal, I immediately think people are dumb when I hear them mention using closed output AI services. Government protections for explicitly anticompetitive services? What a joke!