Igor Kulakov (MicroFactory)
@ihorbeaver • 10,384 subscribers
Making physical products too cheap to meter
Shorts
Videos
Over the past month, we completely switched to frontier LLMs to control a robot in a customer deployment. By doing this, we have a classic System 1/System 2 architecture, but on steroids. For the high-level reasoning layer, we use an agent based on Opus and Astra, which are many generations more capable than any VLA. For the fast execution layer, we use computer vision code written by the model, which is fast, editable, and reliable. As a bonus, we got in-context learning, which is just a basic feature of LLMs, and continual learning, since whenever a new situation arises, the large model steps in and adapts the pipeline. And it just feels better to be on the right side of the Bitter Lesson, benefiting from the billions being invested in frontier models around the world.
Igor Kulakov (MicroFactory)91,607 görüntüleme • 4 gün önce
We’re shipping an interchangeable gripper. Our main goal, as usual, is to make modern AI tools work in environments where you need reliability. Humanoid hands have their use cases, but it’s hard to pick up a tool firmly and control precisely with them. We made an alternative - an interchangeable gripper. The tool is fixed to the gripper in advance, and the whole hand is swapped. AI is quite good at generating this kind of custom tooling.
Igor Kulakov (MicroFactory)929,988 görüntüleme • 3 ay önce
Model precision test. Trained with 2 minutes examples.
Igor Kulakov (MicroFactory)323,938 görüntüleme • 3 ay önce
Introducing an AI robotic integrator. An agent that handles an entire deployment for MicroFactory end-to-end. Powered by the latest Astra and Opus models, it designs tooling, writes computer vision code, and guides a human operator, taking care of all the boring parts of hardware production. Demos:
Igor Kulakov (MicroFactory)53,336 görüntüleme • 21 gün önce
We need self-replicating factories. Reindustrialization won't happen against economic laws. Manufacturing has to be competitive worldwide. And improving existing factories, even with automation, is not enough. Only 30-40k robots were installed in the US last year, so the pull from existing factories is weak. They’re good. Humanoids are being promised as the solution. But who will build them? Still humans. Humanoids not optimised for self-build. The most practical solution to these problems is a self-replicating factory. It will build humanoids, enable the US reindustrialization, and help colonise other planets as a side product. Who should I talk to? I’m assembling people who think the same way.
Igor Kulakov (MicroFactory)20,372 görüntüleme • 14 gün önce
Here is the full run of MicroFactory, that autonomously assembles a photo frame - a real product sold on Amazon. This is a good example of how to deal with complex tasks, while we do not yet have a large set of robotic data to train a big general ai model. Thread:
Igor Kulakov (MicroFactory)210,535 görüntüleme • 11 ay önce
Strong magnets can make the environment self-resetting.
Igor Kulakov (MicroFactory)96,328 görüntüleme • 4 ay önce
We have news! We created a new robotics model called Loop Model 1. On the zip-tie insertion task, it achieves 20x more throughput per unit of data than "Pi06 + RLT" from Physical Intelligence, a top model for such tasks. It’s the missing piece that makes MicroFactory work, because now deployment becomes so simple and fast that our users can do it themselves.
Igor Kulakov (MicroFactory)77,133 görüntüleme • 5 ay önce
Robotic insight 2: One undocumented feature of VLA models is that they can generalize pretty well from very little data, if you run them on good hardware. Here, with 35 demonstrations (5 mins in total), robot can pick up small pcb from any position and at any rotation angle.
Igor Kulakov (MicroFactory)47,571 görüntüleme • 9 ay önce
Bare imitation learning without robotic pre-training is an unfairly overlooked area. (it doesn't even have a name) When done right, it has the advantages of programming. It’s reliable, predictable, easy to modify and debug. The only difference is that the source of truth is a small dataset instead of code. And it is glued together by a vision encoder and a transformer model instead of being compiled. It's hard to imagine how Claude Code could appear without a ton of production-grade source code hardened over years of real use. Same for robotic models. Here is the robustness of such a model. It was trained on orange zip ties with no lighting changes.
Igor Kulakov (MicroFactory)20,322 görüntüleme • 3 ay önce
How to keep your fingers clean while eating chips? Use a robot! - $250 opensource robot with most advanced AI models inside (Tesla bot and Figure use them too) - 15 minutes of examples - $3 for training in Google Collab - Don't hesitate to give a star to the repo if it's interesting:
Igor Kulakov80,442 görüntüleme • 1 yıl önce
Below is our open-source arm assembled by one of the discord members! Also, today Physical Intelligence shared their incredible top-notch robotics model Pi0. And I just want to remind you that you can use our arm together with this model and with LeRobot The arm has the same DOF as the arm used by Pi0, but has smaller size and cost starts from 200$ per arm. You can use better motors like we did in our demos, then it will be more precise and stable but the price will be about $600 per arm. Printable files, purchase list, and help with the assembly - in our discord channel.
Igor Kulakov (MicroFactory)53,938 görüntüleme • 1 yıl önce