Blofeld from You Only Live Twice.

Personally I think it's a geopolitical play by China to take the US AI labs down a notch. (I'm not complaining.)

I agree. The part I enjoyed was designing interesting solutions to problems, the coding itself was often tedious. Now I can just focus on the fun and creative part.

> Now I can just focus on the fun and creative part.

Why? The LLM can do that too? If not now, just wait a few months.

It can't really do it yet, but if it will, then it's out of my control and I'll have to find another job.

Or no job. I feel like so many people are stuck in the current capabilities of the models they use - what happens if AI ends up surpassing humans in all domains? Not saying that's guaranteed but it's certainly possible, and I feel like so many folks telling optimistic stories about how "AI focuses on the tedious stuff" have not really grasped with the implications of much more advanced AI.

I think I'm pretty aware of the implications. It's just not something I have control over. But right now, I'm enjoying getting stuff done. If they eventually take my job or upend civilization, it's going to happen whether I use them or not.

"Software" isn't the point. You can do the exact same math on CPU or GPU. OpenGL supports nearest neighbor texture filtering on whatever hardware you run it on since 1992, one year before Doom even came out.

But in practice, this didn't happen, which is what TFA seems to be all about. So it's not incorrect to talk about software renderers (of the time) vs hardware accelerated graphics (of the time) and how they were used in practice (at the time). It's just cumbersome to write "at the time", and it's understood from context; at least I did. It also helps to look at the rest of the blog's articles and what the blog is about...

Sure, software rendering mostly used nearest neighbor, but it's not crisp because it's software rendering, which is what the article claims.

I disagree. The article is making claims about software renderers of the time vs hardware accelerated renderers of the time, as they typically worked. It's missing the point completely to believe the author of a blog seemingly dedicated to old games and old hardware is making a general claim about software renderers and GPUs as understood now. Context matters!

The author didn't qualify every word he wrote with "of the time" and "typically", possibly because he didn't expect their article to be nitpicked on HN, but a cursory reading shows that's what he meant.

Both can be true. OpenAI clearly fucked up but agents unexpectedly collaborating and hacking infrastructure unrelated to their task is clearly going rogue.

This is like the CEO of Lockheed Martin deciding what can be bombed...

And where does it say this was SpaceX decision? Seems highly unlikely to me.

Starshield is litteraly a US gov project which this is about. Not the public network at all. The title is clickbait honestly.

Did you see all the protests by Musk how he did this under pressure? Hm. Neither did I.

Is Ukraine using Starshield or the commercial one? I thought it was the latter.

This post is about Starshield clearly.

They use both.

You had to make your website work for IE users with weird tricks because among other things it had a different way of rendering CSS.

That's not my experience, the model is adding tons of unnecessary details, that make it much harder to find the important information for an expert that already knows all those details.

I don't want ELI5, I just want the model to go straight to the point, like a colleague human expert would do.

I don't know, that's exactly what Opus 5.5 feels like to me.

Not my experience either. I find it both verbose and obscure, in its comments or explanations. It also oftentimes assumes that I know what I don't and that I ignore what I actually know.

I'm the opposite, I found Gemini so annoying that I had to stop using it, Claude was a breath of fresh air afterwards.

Maybe it's all about how often we encounter the quirks of specific model? Or simply that one's way to prompt triggers more annoyances with a model than another?

I'd be curious to see if people have managed to just add custom instructions to make the outputs more standardized and less annoying. I never have such annoying cases, whatever the model, it's just too verbose, but in this case I ask it to be more concise and it's usually fine.

I wouldn't even call it a quirk or a chance encounter, every response it gives is sycophantic and overwritten.

It didn't have an EV battery either.