Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@zatzi 265 million pounds for toast... and cereal...

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@MagellanQuest Whoa whoa whoa I didn't ask for any of this.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@phyleaux @GatehouseTwenty They do.
Just not at every data center because its immensely more complex and much more equipment to do closed loop cooling vs open loop evaporative cooling.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@vsbuffalo Vibe Researching > Vibe Coding

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@thejelvprint dude...
Bitwise operators are significantly faster than normal math...

Using values that are single bits flipped can be over 100x faster. And is more space effecient as it allows for better inherent alignment...

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@tlbtlbtlb blame openai.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@sparr0 @reset_by_peer Update rot.
The constant change in API for chrome / firefox plugins and android / iOS apps means they need updates to stay functional, and they will also try to update themselves for the same reason.

So unless you're using a non-updating desktop app its just going to rot.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@perrymetzger I didnt say if the code was good quality or not.

Did you miss my point entierly or?

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@g5p8v8r7s7 @RoseSilicon Hydrocarbons are technologically superior to batteries in almost every way. It's 50x more energy dense and less flammable (ironically).

The only improvement would be making synthetic hydrocarbon fuel from renewables / nuclear .

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@zatzi https://t.co/tAVYX1H7Kw
https://t.co/hcTfWqyqS8
https://t.co/iGhopyLSb4
https://t.co/qwGjEfu2pI

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@zatzi People have been quitely working on greening megaprojects across the world, especially in Africa and China. I don't think it has to do with "more CO2", that might kill some plants eventually even.

But it isn't a negative story so it isn't told as often.
https://t.co/qDp3riD4Oh

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@johncrickett This might atleast explain explain why "teams" of agents doesn't seem to scale well. Because they can't build OR mental models of each other. So communication is sort of impossible because they're talking AT either other rather than collaboraitng in shared mental space.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@johncrickett It literally is.
You build a mental model of how the person LLM works then try to predict the output.

Both are equally hard currently. Humans are many a bit easier because the result is relatively consistent within 1 person so I can get used to them. LLMs change inexplicibly.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@cmuratori Superscalar execution is a foreign concept when people talk about performance...

sigh.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@johncrickett Perhaps it's easier to think of them as a pseudo-person, a worker you can command and expect them to do *something*.

Or a mischevious magic hat.
Here's a random thread with the Grok's thought vs my own if you care (it's very long)
https://t.co/2xpl9iJW1p

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@johncrickett I can think of it as helpful, if the situation is right in the same way I can ask customer service for help and how they have the right capabilities.

I can't use a customer service worker as a tool in the traditional sense- but I can lay them among my tools, like a wishing well.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@DigitalJuniors Differences being? Elaborate on what those are and how improve the experience for me if you could.

[Mallchad 🏴󠁧󠁢󠁥󠁮󠁧󠁿 @Mallchad]

@wingedPigBBQ @saleskhalifa This is even before taking into the consideration most chatbots have an pay-per-token interface that you can plug-in to seperate tool using ecosystems... And that many very good models are semi-open source or self-hosted.

The whole statement is either ignorant or dishonest