lousken

lm studio bionic, unsloth, now this... it would be nice if it worked at least in one of those without installing another component

The sooner it pops the better

Since about a month ago, google went to hell. Even bing might be better at this point.

export of the best machines should be banned before europe builds its own chip manufacturing and supply chain

Cost to Run Artificial Analysis Intelligence Index is higher than previous Opus, so still not cheaper

Why do people still use microslop in the age of ai?

Davinci oficially supports Linux but in the worst way possible - "it kinda works if you hack around it, we haven't bothered" There is no official Ubuntu or current RHEL support and you have to mess with libraries and stuff to make it kinda work. No proper window decoderations either

AI companies should pay billions to wayback machine for access

Maybe they should have shorter and clearer terms? Maybe it should be easy to just opt out?

I call BS on that, it's been 6 months since Opus 4.6 and we still do not see a revolutionary software release or a major increase in quality. At least we see some security bugs being finally fixed.

I think GPT 4 was meant to do that.

Can they support something like modern linux? Every single youtuber talks sh/t about davinci and they still do not care

How to view the previous version to compare?

Unfortunately GLM5.3 is a cheaper replacement for haiku than sonnet

While haiku is almost one year old. What a joke

that reminds me to buy some, anyone tried it already?

No they do not. In some tasks yes, but they do lack when producing output in other languages.

who cares about c&d? just host it outside of us

gpt oss killer? this can easily run on a server cpu with its memory bandwidth

Exactly, they need to push to regular consumers as well as businesses. This will create more pressure for adoption

What is a tight leash? Human developer and human reviewer? Or is AI reviewer good enough? ... Couple years ago people were copying code from other parts of a codebase or stackoverflow, now they copy it from some blackbox AI.

Do we need a special way of calling it? How did we call people using stackoverflow?

Are we testing the model or the harness? If benchmarks show certain numbers it should perform as such without it

So it's not vibe coding anymore. Now it is agentic engineering. I just can't...

And that includes gtk3 bloat I am assuming, gtk2 would be even better. But 30MB for a weather app sounds acceptable to me.

I am pretty sure xfce4 weather plugin takes a lot less.

If you use a wrong metric you get wrong results.

Do I still own the device if I can't remove what I want?

The sooner it pops, the better. Later on it will be even worse.

I had been using ai since gpt3.5 and after so many years of bullshit I am also sick and tired of everything related to it. First ai push at work was dumb, now i feel like second push is coming and I do not want to go through slop again. Internet being full of it does not help.

I am surprised they don't record a decrease, maybe this year could be better?