Why is it so hard to build robots?

Mass production now has the price of an embedded computer down to $4 in the case of the Raspberry Pi Pico, however the prices of motors, sensors, and hardware aren't subject to Moore's law in the same way. This means that to personally experiment with robots, you have to spend a boatload of money. Building a robot arm, for example, requires about 8 axis of motion, force feedback, and a fair bit math to even pick up an object and move it to a new location.

The software for robots doesn't really do imitation, or AGI style learning, so you need programming to get it to perform any given task.

We're in robotics about the same place personal computing was in 1975. Some people have made interesting things, but there's only a few commodity systems that work and no real killer application that would deliver real value in the home.

Sure, Boston Dynamics is interesting, but $75000 for Spot is way more than the $500 IMSAI inflation adjusted to about $2500.

Less hackers have access to them, less companies can afford to experiment with them... prices still need to drop a few more orders of magnitude.

I'd say that there are two categories of reasons.

First, the iteration cycle is much longer than for software.

Second, Moravec's Paradox https://en.wikipedia.org/wiki/Moravec%27s_paradox

Both of these make robotics really expensive to research and develop.

Robots like order and predictability. As engineers we try to design robots to meet certain needs. Real life is messy.