rossant

The difference is that bots go much faster and don't take any break. That's in part what makes them so dangerous.

There’s an argument I don’t hear very often. Society is becoming increasingly dependent on data/compute centers running AI models. It would take only a few bombs from a hostile country to disrupt the entire economy, army and so on, especially if workers can no longer do anything useful without AI.

I had to read it 5 times. What was announced exactly? Obfuscation is a huge red flag in announcements like this.

Unpopular opinion but I kind of agree. I've been using Codex extensively these past few months. Using the latest paid models. Something is changing. It was a nice assistant before. In the past few weeks, I realized it became better than me in making architectural decisions on low-level performance-critical code. For the first time, I realize that I prefer his decisions rather than mine. He is just better now. I have 30+ years of programming experience. I agree with the author that we should get ready to adapt quickly to a world that may soon become radically different.

(2007)

I also sense an LLM vibe in this post.

I opened the comments fully expecting the top reply to be “In mice.” Bingo.

Genuine question: how do you backup (and reimport) a bunch of ChatGPT conversations?

Fifteen years ago, I ran a YouTube channel with hundreds of obscure French videos about pediatrics and parenting. One of them suddenly attracted massive attention worldwide, especially from Pakistan and Indonesia. According to the stats, 99% of the viewers were male. Millions and millions of views. For months, it sat in the top five French videos on YouTube. Ad revenue went through the roof, like three figures per day, for months, from that single video. None of the others on the channel saw anything remotely similar. It was baffling.

Then I understood why. The automatic thumbnail generator had picked a frame from the exact middle of the two-minute video. It showed a close-up of a newborn heel prick test: a nurse firmly holding the baby’s heel and pricking it to collect a drop of blood for routine postnatal genetic screening. The thumbnail frame looked like a skin-colored cylinder grasped by a woman’s hand.

Thankfully, the flood of comments, expressing disgust and horror at a medical procedure on a newborn after viewers had expected something entirely different, did not prevent the algorithm from enthusiastically recommending that thumbnail to a significant fraction of humanity.

Zotero had been “not good enough” for many years. Since version 7, it has become nearly perfect for me. I can’t recommend it highly enough.

Out of curiosity, what was the stuff you thought was healthy but was destroying your health?

I do it, but I also like to use headers.

I like to use comments extensively, even if it’s just to visually separate blocks of code in my IDE. A bit more spacing simply feels cleaner to me.

Guilty.

Perhaps closer to 100x actually.

Why can't the LLM refrain from improving a sentence that's already really good? Sometimes I wish the LLM would just tell me, "You asked me to improve this sentence, but it's already great and I don't see anything to change. Any 'improvement' would actually make it worse. Are you sure you want to continue?"

Awesome. This is the way.

Nice, that swaps odd and even around.

Lately, ChatGPT 5.1 has been less guilty of this and sometimes holds off answering fully and just asks me to clarify what I meant.

> As a side note: replacing the chip took longer than expected. I accidentally ordered a GD32F350R8T6, instead of the GD32F350RBT6 that was in the device originally. These two types differ in their flash sizes: 64 kB vs 128 kB. Don’t ask me why GigaDevice thought this naming scheme and this font was a good idea

An 8 looking almost exactly like a B. What a terrible idea.

Very interesting. Is there any other simulation that also exhibits spontaneous illegal activity?

Thanks for this! I wasn't aware of this reference.

I had a great experience with using Hugo to set up a relatively complex multilang website with a lot of custom advanced workflows.

The twist is, OpenAI Codex did all the setup. I described precisely what I wanted in natural language. Codex did a perfect job.

A major milestone in the fight against wrongful medical determinations of SBS/AHT. Proud to have made a tiny contribution to this case (disclosure: signed an amicus curiae). More context: https://news.ycombinator.com/item?id=37650402

That doesn't seem totally implausible.

Have you tried OpenAI Codex with GPT5.1? I'm using it for similar GPU rendering stuff and it appears to do an excellent job.

What prevents LLM designers from cheating and including a human handcrafted SVG into the model for this specific request (allowing for variations between calls)?

Is the idea that ffmpeg could change its license and wreak havoc?

Sure, though there’s a difference between extracting energy stored for millions of years and capturing the continuous flow of energy from the sun.

Sounds like something is going really bad. I'd be very curious to know what's happening.