Michael Mentele
@mmentele.bsky.social
Serial founder. Interested in embodied AI. Top 1% longevity athlete. In it for team human. Likes shiba inus and dnd.
founder https://spaces.sh
co-founder https://leftlanesoftware.com
xr techlead https://www.meta.com/
founder https://spaces.sh
co-founder https://leftlanesoftware.com
xr techlead https://www.meta.com/
I REALLY want to incorporate this into the twin so that you can train robots against your specific twin before deploying them 🤩
Paper: jianghanxiao.github.io/phystwin-web/
Paper: jianghanxiao.github.io/phystwin-web/
October 18, 2025 at 3:34 PM
I REALLY want to incorporate this into the twin so that you can train robots against your specific twin before deploying them 🤩
Paper: jianghanxiao.github.io/phystwin-web/
Paper: jianghanxiao.github.io/phystwin-web/
US is 24% of global GDP but 40% of capital markets. Does that make sense?
October 17, 2025 at 8:03 PM
US is 24% of global GDP but 40% of capital markets. Does that make sense?
I hate the state of IoT with a burning passion. Everything is locked down. Just give me an API like this. I'm about to make my own 2-way audio device that isn't stupid.
October 17, 2025 at 3:34 PM
I hate the state of IoT with a burning passion. Everything is locked down. Just give me an API like this. I'm about to make my own 2-way audio device that isn't stupid.
Rigid robots -> tight tolerances -> classical sim + teleop 👎
This isn't the way, it doesn't scale, and it doesn't generalize.
This isn't the way, it doesn't scale, and it doesn't generalize.
October 16, 2025 at 8:02 PM
Rigid robots -> tight tolerances -> classical sim + teleop 👎
This isn't the way, it doesn't scale, and it doesn't generalize.
This isn't the way, it doesn't scale, and it doesn't generalize.
The limiting factor for real AI is data. <0.001% of the worlds information is on the internet. What happens when machines get access to the rest?
October 16, 2025 at 3:34 PM
The limiting factor for real AI is data. <0.001% of the worlds information is on the internet. What happens when machines get access to the rest?
K-Scale has a few awesome videos breaking down their hardware and real2sim methods -- recommend checking them out. All their stuff is open source ❤️
www.kscale.dev/kbot
www.kscale.dev/kbot
October 15, 2025 at 8:02 PM
K-Scale has a few awesome videos breaking down their hardware and real2sim methods -- recommend checking them out. All their stuff is open source ❤️
www.kscale.dev/kbot
www.kscale.dev/kbot
Slow down for demos while going through MIT's graduate computer vision course. Learning about NERFs, diffusion, splats, optical flow, and other magical topics.
www.scenerepresentations.org/courses/2025...
They have tons of open courses: ocw.mit.edu. Been using it for years.
I ❤️ you MIT.
www.scenerepresentations.org/courses/2025...
They have tons of open courses: ocw.mit.edu. Been using it for years.
I ❤️ you MIT.
October 15, 2025 at 3:34 PM
Slow down for demos while going through MIT's graduate computer vision course. Learning about NERFs, diffusion, splats, optical flow, and other magical topics.
www.scenerepresentations.org/courses/2025...
They have tons of open courses: ocw.mit.edu. Been using it for years.
I ❤️ you MIT.
www.scenerepresentations.org/courses/2025...
They have tons of open courses: ocw.mit.edu. Been using it for years.
I ❤️ you MIT.
First integration test for pose projection into a twin.
The nice thing is this works for any kind of pose, not just a person. So, I see a clear path to go from from 1 -> 1000 object types with potentially just a few weeks of effort.
The nice thing is this works for any kind of pose, not just a person. So, I see a clear path to go from from 1 -> 1000 object types with potentially just a few weeks of effort.
September 30, 2025 at 8:02 PM
First integration test for pose projection into a twin.
The nice thing is this works for any kind of pose, not just a person. So, I see a clear path to go from from 1 -> 1000 object types with potentially just a few weeks of effort.
The nice thing is this works for any kind of pose, not just a person. So, I see a clear path to go from from 1 -> 1000 object types with potentially just a few weeks of effort.
From a first order POV generating digital stuff is cute and fun, but doesn't stop Chinese supremacy or make real life observably better.
September 30, 2025 at 3:34 PM
From a first order POV generating digital stuff is cute and fun, but doesn't stop Chinese supremacy or make real life observably better.
I had to rework how I was projecting poses in real time. First baby step on the renderer side.
At some point I want to project a realistic 3D mesh but for now simple 3D pose skeletons should get us 80% of the way.
At some point I want to project a realistic 3D mesh but for now simple 3D pose skeletons should get us 80% of the way.
September 24, 2025 at 8:03 PM
I had to rework how I was projecting poses in real time. First baby step on the renderer side.
At some point I want to project a realistic 3D mesh but for now simple 3D pose skeletons should get us 80% of the way.
At some point I want to project a realistic 3D mesh but for now simple 3D pose skeletons should get us 80% of the way.
We suck at STEM. And aren't getting better... thank god for imports keeping us afloat. IDK how long we can keep the inertia going tho.
September 12, 2025 at 8:02 PM
We suck at STEM. And aren't getting better... thank god for imports keeping us afloat. IDK how long we can keep the inertia going tho.
You make polish with friction and grinding. Not pampering and hand holding.
September 12, 2025 at 3:34 PM
You make polish with friction and grinding. Not pampering and hand holding.
Realtime 3D pose projection, next step wiring into the twin for telekinesis-like super powers.
September 12, 2025 at 1:17 PM
Realtime 3D pose projection, next step wiring into the twin for telekinesis-like super powers.
This week working on pose capture and projection into Space. Once that's done we should be into application land then onto the SDK 🤘
September 12, 2025 at 1:16 PM
This week working on pose capture and projection into Space. Once that's done we should be into application land then onto the SDK 🤘
If we don't learn to make real sh*t we are f'd.
September 11, 2025 at 8:02 PM
If we don't learn to make real sh*t we are f'd.
In Thailand at 22 I ran out of money and a temple took me in. I lived with the monks for a month meditating 6 hours a day. It was life changing.
September 11, 2025 at 3:34 PM
In Thailand at 22 I ran out of money and a temple took me in. I lived with the monks for a month meditating 6 hours a day. It was life changing.
I've run 11 miles on a broken foot and 3 on two...
But I didn't care because my mindset was I will finish or die. Perhaps myopic (haha, I couldn't stand for weeks) but true commitment makes hard or easy irrelevant.
It's actually relaxing.
But I didn't care because my mindset was I will finish or die. Perhaps myopic (haha, I couldn't stand for weeks) but true commitment makes hard or easy irrelevant.
It's actually relaxing.
September 10, 2025 at 8:02 PM
I've run 11 miles on a broken foot and 3 on two...
But I didn't care because my mindset was I will finish or die. Perhaps myopic (haha, I couldn't stand for weeks) but true commitment makes hard or easy irrelevant.
It's actually relaxing.
But I didn't care because my mindset was I will finish or die. Perhaps myopic (haha, I couldn't stand for weeks) but true commitment makes hard or easy irrelevant.
It's actually relaxing.
I used to hate incompetence but then I realized it's just my opportunity to be superhuman.
September 10, 2025 at 3:34 PM
I used to hate incompetence but then I realized it's just my opportunity to be superhuman.
Stop hijacking dopamine. Start making robots, 3D printers, manufacturing lines, sensors, drones, bridges, trains, transit systems... things that actually matter.
September 9, 2025 at 8:02 PM
Stop hijacking dopamine. Start making robots, 3D printers, manufacturing lines, sensors, drones, bridges, trains, transit systems... things that actually matter.
Them: technology is advancing so fast!
Me: <looks around> I still live in a mud and stick house and use four wheels to drive on big flat rocks and stare at 2D screens <scratch head>
Me: <looks around> I still live in a mud and stick house and use four wheels to drive on big flat rocks and stare at 2D screens <scratch head>
September 9, 2025 at 3:34 PM
Them: technology is advancing so fast!
Me: <looks around> I still live in a mud and stick house and use four wheels to drive on big flat rocks and stare at 2D screens <scratch head>
Me: <looks around> I still live in a mud and stick house and use four wheels to drive on big flat rocks and stare at 2D screens <scratch head>
Delete cheap dopamine: games, tv, porn, etc. You'll normalize to higher protein options and have just as much fun.
September 8, 2025 at 8:01 PM
Delete cheap dopamine: games, tv, porn, etc. You'll normalize to higher protein options and have just as much fun.
Turkey's will sit on a football if you make it chirp. Video games do the same thing for your progress instinct.
September 8, 2025 at 3:34 PM
Turkey's will sit on a football if you make it chirp. Video games do the same thing for your progress instinct.
Thoughts on this style of API for getting spatial events out of Space?
September 7, 2025 at 3:34 PM
Thoughts on this style of API for getting spatial events out of Space?
This week hasn't had many visuals for Space, mostly architecture planning -- but we should close the loop on the full system in the next few weeks.
September 6, 2025 at 8:01 PM
This week hasn't had many visuals for Space, mostly architecture planning -- but we should close the loop on the full system in the next few weeks.
Zones now spawn where you are looking (before they always spawned at the origin lol).
September 6, 2025 at 3:34 PM
Zones now spawn where you are looking (before they always spawned at the origin lol).