Ive taken too much coffee and am consumed by posting, but theres legit so much space for automation. One of my favourite was a simple small crane whose sole job was just lifting solar panels for commercial install and it like 5x's productivity
Replies
Hell yeah brother youtu.be/ZZ2fP1Y5Z2E?...
aug 31, 2025, 3:27 pm • 4 1 • view
its so sick lol.
aug 31, 2025, 3:35 pm • 2 0 • view
a robot that can install framework panels and support posts for them, at heights of up to 5.7m - with no extra scaffolding www.youtube.com/watch?v=Akl6...
aug 31, 2025, 5:30 pm • 0 0 • view
Boy there is a whole book waiting to be written on how the blind spots in the experience of the tech crowd leads to massive misses in development. They create Uber and DoorDash and LLMs because they address problems the creators see every day. But there are tons of other opportunities out there.
aug 31, 2025, 3:31 pm • 2 0 • view
New theory for why europe might do well is that now that we are out this digital gold rush, traditional funding models make more sense again
aug 31, 2025, 3:34 pm • 3 0 • view
i basically understand this feature of the US as being because of how the funding system works
aug 31, 2025, 3:38 pm • 1 0 • view
(sobbing) robot please just pull my weeds.
aug 31, 2025, 3:27 pm • 2 0 • view
Training a small little robot with an arm to kill this fucking goutweed
aug 31, 2025, 3:29 pm • 1 0 • view
That came from one of the commercial installers on here; but also other robotics people are posting about tiny arms to do stuff like small scale packaging and i think we are getting to a point in the next 5 years where you can just tell an LLM what you want, and it can roughly figure it out
aug 31, 2025, 3:27 pm • 9 0 • view
This is sort of true, but the sheer difficulty of creating enough training data might prevent it from being good enough to scale. @cpaxton.bsky.social has written a lot about this. e.g. itcanthink.substack.com/p/how-can-we...
aug 31, 2025, 3:43 pm • 5 0 • view
You don't need robot GPT for this though. These agents do specific, well-defined tasks in a domain that can be curated to make it easier for them. This is exactly the sort of thing you can simulate the heck out of
aug 31, 2025, 3:44 pm • 5 0 • view
I agree that’s how to actually scale robotics, but that’s different than “prompt a common robot and it will figure out how to do it.”
aug 31, 2025, 3:53 pm • 1 0 • view
right i'm more thinking of some control scheme, where given arbitrary inputs and outputs, some kind of robot can figure it out so we can automate a lot of what stops us from scaling it. No where close to that rn
aug 31, 2025, 4:08 pm • 0 0 • view
domain specialization + teleoperation data
aug 31, 2025, 4:50 pm • 1 0 • view
5 years would be remarkably fast.
aug 31, 2025, 3:44 pm • 2 0 • view
(I just say this because auto maker had some investments in this space and was not expecting pay off that quick)
aug 31, 2025, 3:45 pm • 3 0 • view
its going to ofc be some grey area where it works on some things but not on others, but we are still pretty far from that
aug 31, 2025, 4:09 pm • 1 0 • view
(obviously not that complex but you know, lift thing to place, make some boxes, these kinds of things)
aug 31, 2025, 3:27 pm • 4 0 • view
Ugh i should be working on this probably
aug 31, 2025, 3:30 pm • 3 0 • view
The intersection of AI and robotics is super interesting, especially with more modular robotic designs, programmable "brains", and relatively affordable costs. Semi autonomous robots custom assembled and programmed for specific tasks/niches is really worth looking into. But US isn't the place.
aug 31, 2025, 4:29 pm • 2 0 • view
This is the ideal robot body. You may not like it, but this is what peak performance looks like. www.hestia.kitchen/multicooker
aug 31, 2025, 4:33 pm • 3 1 • view
Ooh there's an animation if you click through!
aug 31, 2025, 4:36 pm • 0 0 • view