Connection Systems: From a Snap to a Policy
The bridge from a maker's snap-together kit to a learned policy on real hardware. Build the sense-think-act loop in code, read a sensor over a real bus, watch a block graph become a policy that out-learns your hand rule, switch a magnet that holds for free (the energy a body already has can be the guarantee, the seed of the certificate idea) and clone an expert from its own demonstrations.
▶ Start the course ← All coursesThe credential the ecosystem leaves on the table.
The $100 SO-101 arm, Hugging Face's LeRobot, and the worldwide hackathons made teleoperate → record → train → deploy something anyone can do. What they do not give you is a credential, the LeRobot course issues none, by design. This course credentials the same skills, assessed on your own runs: pass the labs and claim a verifiable, standards-aligned certificate you can take anywhere. We certify what you demonstrate here, not a course we did not run.
Where this sits, and what moves it.
Binding constraint · Energy to hold versus energy to switch. A body that must spend power to maintain a state has a runtime; a body whose state is held by its own materials has a mission.
A snap-together kit and a learned policy were different worlds with different audiences. Getting from a block that steers to a policy that learns meant leaving the kit behind and starting over in a research stack.
This course keeps the loop and grows it, ending at a clone trained from its own demonstrations. The electropermanent magnet is the seed of a larger idea, and it is worth being explicit that it is a seed: paying 0.3 J once and holding for free is the same move as a certificate that comes from structure rather than from checking.
Self-reconfiguring machines are waiting on latch energy and alignment tolerance, not on algorithms. Watch joules per connection and capture-basin width -- when those two cross, modular robots stop being demonstrations and start being infrastructure.
Every hard thing was impossible until the constraint that made it impossible was named. How we read a frontier →
The loop, snapped together
See that a snap-together kit is a sense-think-act loop, and that the loop is code you can write.
- L0The loop you can holdYou snap together a sense, think, act graph and then write the think step in code. What is the relationship between the two?Write a sense-think-act controller that drives an agent to a target, and see that the graph you snapped together is a program.→
- L1What the wire carriesYou read a sensor over a two-wire bus and get two bytes. What do you have to know that the bytes do not tell you?Read a real sensor value off a two-wire bus and route it into a decision -- the connector layer, in code.→
When the graph becomes a policy
Cross from a hand-tuned block to a learned one, switch a magnet that holds for free, and clone an expert from its demonstrations.
- L2The graph is a policyThe motor is wired backwards. Your hand rule uses a positive gain to steer toward the target. Does it reach?Watch a learned think block discover a fix your hand rule could not -- the step from maker to physical AI.→
- L1The magnet, switchedOver one minute of holding, which costs less energy?Model an electropermanent magnet: hold force versus gap, and why switching beats holding.→
- L2From teleop to a released policyYou record an expert by teleoperation and clone the policy by least squares. What tells you the clone actually learned the expert?Record demonstrations from an expert, clone a policy by least squares, and deploy the clone.→