Stringman is one of the few robots under $1000 that reach a whole room and lift 2kg. And it's the only cable robot that can run physical intelligence models using @LeRobotHF Use it to ferry items to other robot stations, or just bring yourself a drink. A kid can drive it around with a gamepad like a drone. Pre-order at neufangled.com/stringman. We're ready to ship as soon as @Raspberry_Pi Zero 2Ws are available. There are currently 8 of them out in the wild and we're building a dataset to serve as the foundation of cleaning tasks for cable robots.

Finally reliable self parking

Stringman now orbits the person in the room holding the gamepad so you don't have to hold that little tag up in orbit mode.

When I worked in software companies and had managers, some of the managers didn't know how anything worked and they were terrible. And some of the managers knew the code by heart and had really strong opinions about the architecture and they were great. They made good products. They scoped excellent work for their team. So when I run llms I try to be the second type of manager.

Unless I see a good reason to stop, I can just keep splitting my model apart and getting better results. Why not have a grasping model for dark backgrounds and one for light backgrounds? I only have to load one of them, and each of them works better in it's niche than a generalist.

Am I right to focus on price? Consumers have bought millions of robots under $1000. Cheap robots get deployed, improved, and adopted. They do not have to be perfect or even useful to sell They only need to be affordable and do one thing well. Robot cost is actuator count plus battery size. Throw out the battery and minimize actuators with a CDPR Use external compute and cost and stays within reach of millions of people. Such a bot is still incredibly capable with the right models.

Any other founders waking up early to build a bunch of something in their product?

Did you guys know that there's an awesome room sized 3D printer based on a CDPR called Hangprinter by @tobbelobb and it's got probably 10x the precision of Stringman hangprinter.org/

Several of my suppliers are just fishing shops that have no idea they're part of the American robotics revolution.

If you are an interesting person, you need a crane robot. You already have walls, so you're halfway there! neufangled.com/stringman

Imagine how many more people could design a PCB If there was a self-contained power supply part you could just slap on it and never worry about again. Perfect stable DC at any voltages you want. I just wanted to connect some sensors to an MC okay. I didn't ever want to learn about brownouts and tank circuits and TSVs

It's a common marketing exercise to make personas, so here's the breakdown. 66% - Tyler Millenial. Angry and going broke. Has plenty of problems but your product cannot solve them. scraping by, as long as they don't buy anything non-essential. 33% - Steve Boomer. Bored and wealthy. Has no problems to solve. Would like to enjoy three more nice vacations before dying. 1% - Alexander X. Likes tech, tries new things. Very intelligent. Building a lab and has problems to solve stemming from a desire to do something bizarre. Plenty of money.

I fought with some tiny ribbon cables, but this one was an insult

It looks like there is already a simple solution to mounting heavy objects in drop ceilings. replace an entire tile. I just don't know if this can handle the large lateral loads that Stringman would cause.

Moondream can locate targets out of the box, but the quality of results are no better than the model I've trained. :(

Good grief. Don't talk to customers or they'll convince you that you should have started a militia instead of a company.