l'empathie mécanique @dpwiz@qoto.org

@tristanC Interesting.. My parts catalog only lists knobs with detents that have 32 or less steps...

And I'm fiddling with my nanokontrol knobs with 127 steps in less than a turn and find the resolution abysmal (for camera control).

I don't know about your case, but I think I will be going all in on resolution and normalize the range myself.

@wilfredh HTTP3 cometh!

(IPv6 enables wider QUIC adoption by making UDP hole-punching and the other connectivity hacks unnecessary)

@dpiponi literally unplayable

Well, thanks, I guess...

@kravietz oh no.

I think they made the system a huge disservice by working with the "use our library" instead of publishing wire protocol and inviting multiple implementations.

@DLC @ireneista WDYM "never happened"? Shell navigation and tool-use is a staple skill for any self-respecting model.

@DLC @ireneista IDK, I think Siri would be able to replicate this level of performance

@carnage4life 1. OpenAI et al also have a huge revenues that they can capture if/when the need comes.
2 is a nothingsburger. Enron was just a scumbag company. If you position Lucent as Nvidia then... see point 1. There is already too much cash on their hands, what are they supposed to do with it, buy bonds?
3. we'll see 🤷‍♀️ Don't bet with what you can't afford to lose.

:syntax on

@gregeganSF BTW, since Grok is unicode-aware it is trivial to elicit answers of arbitrary unhingedness with prompts without an apparent setup.
The public then will take screenshots or otherwise edit away the trigger by accident or for cover.

@th standardized and linguistically neutral abomination

@neil This line of argument is self-contradicting. There is a vice loop in it: if the people are losing skills they would delegate even more to compensate, fueling the demand.

@puffnfresh @redmp Yes. Should come together with NoPositiveLiterals

@zeux Being an Apple dev is hard

@lproven @mike It will help making a better non-AGI product. While not AGI (good!) they can be useful. And that direction is already profitable (opex-wise, if you stop doing research and scaling). It is financially prudent to expand.
Maybe the promise to VCs are outrageous, but the product itself is solid (and will benefit greatly from bigger buildouts).

I hate LLMs and the crap they sometimes unload on my screen. They're neither a good product at the time, nor seeds for the superintelligence. But the "nothing more" part is just objectively false.

Pyramid schemes have no value beyond pyramiding and scheming. LLMs have the value of arguably better search engines. People are paying real cash to get real skills and tutoring out of them.

I get that "imperial" sounds majestic.
But how it is even possible to do science using subjective units "like ass-warm per square pint"?
Besides sociology ofc

@cfbolz @DRMacIver Python is even easier to learn if you `dis` a lot and know how does the magic work. Don't understand a concept? dis it! Can't figure out the magic? dis it!

Your OKRs are my MTBF