You had me at 8 times the speed and you lost me at 6 times the cost. I think I can wait a few seconds.

Claude Harness. :’)

“Our friends at HuggingFace” is next-level trolling. xD

😂“Our friends at HuggingFace[…]” who didn’t sue us after our agents ripped them apart because they admitted to not having enough money for legal fees and were in the middle of being bought by nvidia, one of our main stakeholders.

I first thought that guy coming from right side drove on a bike. He glided towards his final position in such a smooth manner 😁

i can't stop hearing "daddy" 😭😭

this is one of the reasons why I use openai. They let you use their other products without using your usage. Anthropic could learn from this.

holy script

you know what feature I want them to add back into the ChatGPT website... The ability to right click on new chat and open in new tab.

lol how petty.

“twitter”

wait whatt?

it didn't generate the bar chart lol

- The IRONY... Asking for a HUMAN REVIEW for an AI Plugin for Developers Submitting Code Created By AI, and Help Support That is using AI swamped by support tickets.

there goes confluence🗑

that’s gonna leave a mark

there goes Jev🗑

= forced clapping >_

best part of the dev day

bro it’s the TELEPROMPTER. Everyone keeps tripping over their words after looking directly at the camera, even this confident extremely suave gentlemen here

That's good and all, because it's an already existing knowledge. But If it is something emergent something that requires not just choosing but making new variables and using that as choise etc, then using multiple transformes just won't work as well overtime. It just doesn't. Yall basically are splitting the system that makes it actually intelligent so that you get get accuracy easily. The day yall can build the same thing on a single transformer OR y'all fineture your latest models atleast every 15 days, It's pointless... If you can make it using a single connection, it could last a lot more than just 2 weeks... Assuming that's the time non planned reasoning stays grounded in usual ai models... Edit: BTW I get that theoretically y'all might think the different connections just train different perspectives of the same data, BUT NO. It may look like like it's training on a specific aspect only, it just doesn't work that way practically. Most of it works as planned but some parts that don't is what actually likely bring the emergent intelligence.

Just use Openclaw / Hermes... As soon as local hosting of open weight models become a lot more cheaper, likely after physical AI becomes normalised, which will change prices of physical things ( not rarity - that's intelligence based ) which will probably either let everyone have an intelligence fridge OR AI t@keover, OR some human tries to do the same but eventually AI t@keover...

OK, I don't want to be that guy who talks all theoretically about tech and algorithms. So I want to try give a better explanation. You know we can reason in our minds and we can store information in our minds right? Now some people can store a lot of information in their mind. They can use that information to find new information but it takes a long time. But others who do the same but through relations, can do a lot easier job and remembers the same things a lot better. BUT the hardest way to do this is to learn the information the exact same way the information came to form. Which likely won't make much sense at first when you do it, but in every circumstance that Information may be of use to you, you will remember it. Because the path the information took to come of existence was the same or very similiar to the one you took to learn it. When you just keep on adding different connections that focus on different aspects and just different connections in general, you basically are memorizing everything using small connections that doesn't actually have anything to do with that information other than you.