Mikhail Liublin

software - hardware - poker - software - ai - robots

[Mikhail Liublin @mlcka3i]

@markgalkev FTW!!!

[Mikhail Liublin @mlcka3i]

@Rainmaker1973 everyone's making warehouse robots. the real moat isn't the hardware — it's showing up to a client site, not stopping the line, and having it actually work on day one. most can't do that.

[Mikhail Liublin @mlcka3i]

@Saint_n0mad cycloidal reducers 3D printed from scratch? the backlash tolerance on those things is brutal. props for getting 8 encoders and ROS integration running on a scratch build. this is the stuff that looks easy until you try it.

[Mikhail Liublin @mlcka3i]

@IlirAliu_ demos get infinite retries and perfect lighting. production gets dropped connections, spilled oil, and operators who don't read manuals. the companies winning are the ones honest about that gap upfront

[Mikhail Liublin @mlcka3i]

Tesla building Optimus lines for 10M units, Pudu going industrial, Accenture running real warehouse pilots. Not just demos anymore.

We're entering the "actually make it work" phase of humanoid robotics.

[Mikhail Liublin @mlcka3i]

Chef Robotics just quietly hit 100 million meals served.

That's the thing about food service robotics it doesn't go viral like humanoid factory demos, but it's already eating the market. Real kitchens, real customers, real repetition.

The boring use cases usually win first.

[Mikhail Liublin @mlcka3i]

Closed humanoid stacks are done. The race is who builds the best system with pieces nobody owns.

[Mikhail Liublin @mlcka3i]

Boston Dynamics put Gemini in Spot. You tell it "pick up the shoes" instead of writing 200 lines of state machine code and it just... does it.

[Mikhail Liublin @mlcka3i]

@tom_doerr This is the kind of open source robotics we need more of not just the software, but the hardware designs, BOM, control stack. Makes prototyping actually accessible for builders.

[Mikhail Liublin @mlcka3i]

@XRoboHub Every 30 minutes is still impressive for humanoids at this stage. The real story isn't the speed it's that contracts exist for 10K units. Someone's actually buying these in volume.

[Mikhail Liublin @mlcka3i]

@RodmanAi robot OS layer is an inevitable bet. hardware stays messy but the dev experience shouldn't be. If this works consistently across form factors it lowers the barrier significantly.

[Mikhail Liublin @mlcka3i]

@Ravenismeee Giving up is for pussies

[Mikhail Liublin @mlcka3i]

@fadule_ forgot the best one: you can 3D print a drone frame from open source files on a $300 printer and have an autonomous flying machine by the weekend. we're not grateful enough for this timeline

[Mikhail Liublin @mlcka3i]

The hard part isn't fixing clogs or color matching. It's realizing the next level requires business dev, storefront, and scaling skills you might not want to develop.

Sometimes the win is knowing when not to level up.

[Mikhail Liublin @mlcka3i]

Love this: founder runs 3D printing side business for 8 months, makes it profitable, then shuts it down because he "understood what comes next"

https://t.co/eDg8HQap4c

[Mikhail Liublin @mlcka3i]

@amazingthings_ What is this place?

[Mikhail Liublin @mlcka3i]

@sciencegirl 3D printing drone frames at this scale and people still tell me milled parts are the only way. The gap closed months ago.

[Mikhail Liublin @mlcka3i]

@YIMBYPoland A humanoid robot named Edward chasing wild boars in Warsaw. I'm not ready for this timeline.

[Mikhail Liublin @mlcka3i]

@spaceandtech_ Shenzhen startup put a new brain in a Unitree G1. Now it waters plants and cleans autonomously. Open source hardware + smart software = moving fast

[Mikhail Liublin @mlcka3i]

@DrJimFan RL-trained robot walking like a human. Trained in simulation, deployed in reality. Sim2real is quietly becoming the default path for robot locomotion.