Yes, dipshits like that should be banned from the medical profession.

Repo level markdown files with architectural guidance not working for example… I’ve found that works quite well.

Same. AGENTS.md files and the like are quite effective. Especially if you’re reviewing the code and making the LLM help you update the markdown files when it makes a mistake to prevent the same type of mistake in the future. Having concrete examples of “good” vs. “bad” to illustrate each architectural rule goes a long way.

For any feature or bug fix that is “painting with the colors already in the tray”, it makes sense to let a LLM write the code. Humans will introduce new tech and new patterns out of boredom and turn the codebase into a big Frankenstein, but the LLM will just follow the architectural guidelines indefinitely.

Children often just want sweets and unhealthy snacks, and it’s up to adults to make sure they get the nutrition their little bodies require. If a grown ass adult decides they don’t want to eat something, though, they likely have tried it and have their reasons for avoiding it. It’s really nobody else’s damned business.

It will have strong privacy and security options so you can trust it to handle all of your personal content knowing that no one else can access your information, similar to how encryption works on WhatsApp

Judging by how “private” WhatsApp is, no fucking thank you.

Qwen 3.6 27B has similar capabilities to Claude Sonnet 4.6 for coding, but uses about as much power for inference as playing a game on the gaming PC running it, compared to Sonnet that probably runs on like 8 B200s in a data center. If people were mostly just running open weight models on gaming PCs with no tech bros involved, I don’t think we’d be seeing the same level of outrage.

If the Meta devs find their work drying up as a result and are risking a layoff, they could always contribute anonymously to keep the racket going perpetually.

Agreed, this is one way that LLMs can increase the quality of code.

A lot of times, you just get a big bill without them doing anything useful for you.

Wow, nice. It sounds like Asus makes it the easiest, reimbursing for the license without shipping your machine to them. I don’t understand why the others that will reimburse require physical inspection of the machine.

It makes sense to me that a software pipeline transforming copyrighted material into new IP should be bound by the terms of the copyright agreements. Unfortunately, these companies can afford to hire cream of the crop legal departments to bend existing laws like fair use in their favor. Unless some strong legal precedents emerge to require that training data respect existing copyright right laws, legislation might be the only hope, but of course, deep pockets also help with influencing legislators.

Maybe they don’t make it like that anymore, but AI will surely be used to pump out a bunch of slop just like it.

no one converts to windows later in life

That’s why you gotta start kids off right. Perfect starter laptop is the 10 year old clunker running Mint.

You just need the right soundboard. This is a personal favorite: https://soundboard.com/sb/angryblackguy1.

Don’t caaall my fuckin’ house

🏆

Greenland is going to live up to its name pretty soon at this rate (any part still above water, that is).

I played a used copy of GTA 5 for a few hours and then got bored and traded it for something else. Since used copies won’t be a thing for GTA 6, I’m sure there are other ways to get a copy to mess around with without wasting a lot of money on it.

There’s a good chance you’re right, unfortunately. I guess copyright laws in general need revamping.

I’m a fan of Rick Beato’s YouTube channel. It’s funny watching him review the Spotify top 10 as someone who is genuinely open-minded and who appreciates all forms of music. He wants to find something to like, but often struggles because they mostly have the same cliche drum beat, are short vamps instead of songs, etc. We already had enough low-effort, human-generated slop without AI taking it to a whole new low.

When there’s a dispute, courts consider the following four issues in deciding whether a use is fair use:

1. why the party used the copyrighted material (for instance, for commercial versus educational purposes)
2. whether the copyrighted work is informational or for entertainment
3. how much of the copyrighted work the party used, and
4. whether and how the use affects the market for or value of the copyrighted work.

1. Commercial
2. Informational
3. All of it
4. Definitely does

Based on this, not sure fair use holds up that well, but I’m not a lawyer.

Sure, train models on copyleft data as long as all the training data and pipeline code to train the model is open sourced under the terms of the license.