I stopped buying humanoid robots — and started printing them

Last March I preordered a 1X NEO. Six months later I cancelled, returned the deposit, and spent a weekend printing the frame of a Poppy humanoid. Here's what that year taught me about the robot I actually wanted.

What the expensive ones get right

NEO walks. It's quiet. It holds my cup of coffee without spilling it. The perception pipeline is magical. The industrial design is Apple-grade. When the 1X team nails a demo, it feels like watching the future.

What they never let you do

• Open the firmware
• Rewrite a gait
• Replace a joint with something better
• Look at the training data
• Extend the robot beyond what the vendor pre-approved

You rent a robot, basically. The warranty evaporates the moment you try anything interesting.

What printing Poppy taught me

Poppy is janky. The SG-90 servos buzz. The gait is closer to "tipping repeatedly in the correct direction" than walking. The BOM is $500, the print time is 80 hours, and the first assembly took a weekend and a YouTube tutorial.

But I own it. I broke a shoulder joint last month and had a replacement printed in three hours. I added a Raspberry Pi 5 for onboard inference. I'm experimenting with a reinforcement-learned gait over the weekend — no legal review, no firmware signing, no vendor portal.

What I actually recommend

If you want a robot that works, buy the closed one. If you want to learn, print the open one. They serve different goals. Most people who ask me "which humanoid should I buy?" actually want the second — they want to mess around — and then end up frustrated that their $20K investment has no hack access.

RobotForge is built for the second crowd. You know who you are.