Joseph

Cooking to replace cooks cooking with robo cooks. lil cook-oo, views are my own

[Joseph @josepager]

@DataScienceHarp Very nice! Though one of the lowest hanging fruits for collection, its one of the most complex and will require immense amounts of data to achieve great execution. Would love to talk. Commercial seems the best way when it come to kitchens

[Joseph @josepager]

@antopatrex1 I think SF wants to hear it lol. Different story is someone putting their rep on the line to do it

[Joseph @josepager]

@stijnnoorman Facts

[Joseph @josepager]

@therealchaseeb 💀💀💀

[Joseph @josepager]

@thedankoe

[DAN KOE @thedankoe]

When calories became cheap, convenient, and abundant, bodies decayed.

Thinking is now cheap.

[Joseph @josepager]

Might've been wrong on my assessment china won't win at WAMs

[Cristóbal Valenzuela @c_valenzuelab]

Heard a crazy stat the other day. Video generation represents around 70% of the share of AI token consumption in China. A combination of short form content and robotics. In China, video models are growing much faster than Claude Code grew in the America.

America is LLM-pilled. China is World-Model-Pilled.

[Cristóbal Valenzuela @c_valenzuelab]

Heard a crazy stat the other day. Video generation represents around 70% of the share of AI token consumption in China. A combination of short form content and robotics. In China, video models are growing much faster than Claude Code grew in the America.

America is LLM-pilled. China is World-Model-Pilled.

[Joseph @josepager]

@lukas_m_ziegler Bet this wasn't on any of the bag holder's bingo cards.

[Joseph @josepager]

@austinxwalker Whole foods and exercise. Keep it simple

[Joseph @josepager]

@sarahookr @_varunnair Right on the money here. We're gonna see a huge boom and bust of data companies trying to fill the void. Only to have a goated sim rug everyone by being able to synthesize the real world.

[Joseph @josepager]

@ClaraChengGo Any hardware providers selling dual kits at scale already?

[Joseph @josepager]

@YouJiacheng Hard with them THICC parameter size

[Joseph @josepager]

@Jeffjinlin Doing this myself in the kitchen! Would love to ping pong ideas!

[Joseph @josepager]

@dannyngwsh Agree, people love to crap on US robotics but Chinese companies are doing great only o creating the proper production hardware pipeline. THey’re far behind on the intelligence side and are shamelessly waiting for the US to solve it for them.

[Joseph @josepager]

@audrlo Don’t think this statement is as bullish as you think it is.

[edgar @edgarpavlovsky]

In a world of slop software, the products that will stand out are those deeply understood and intentionally architected by their builders

the age of the "quick mvp" is over - the quick mvp is now just slop

[Joseph @josepager]

A potential mini Theranos situation dare I say

[TechCrunch @TechCrunch]

The lawsuit alleges that Oura rings are unable to measure any of the physiological signals needed to assess sleep quality or determine sleep stages.
https://t.co/JiIB1K4npV

[Anto Patrex @antopatrex1]

nvidia's jetson orin runs a 3B parameter model at 15 tokens per second on 30 watts. qualcomm's robotics platform is pushing similar numbers. on-device inference for robots just became real and the deployment implications are massive.

right now most capable robots are basically thin clients. they run their policy network on a cloud GPU, stream sensor data up, get actions back. works fine in a demo. completely falls apart in a warehouse with spotty wifi or a construction site with no connectivity at all.

your robot can think locally now, react in real time, and doesn't need a $40k cloud GPU bill every month. the economics shift completely when inference is local. you go from needing enterprise level connectivity infrastructure at every deployment site to just needing power and wifi for periodic model updates.

that's the difference between robots only working in amazon warehouses and robots working in a random machine shop in ohio. edge compute is what makes physical AI a small business technology. and that's where 90% of the actual deployment volume is going to come from.