Writing

The decisions, and the ones we got wrong first.

Every piece here comes from a real architectural decision with a real cost attached. The post-mortems are included on purpose: a company that only publishes its wins is telling you what it wants to be true.

Post-mortem6 min

The flywheel, and how it fails silently

Our data loop was wired, tested, and moving nothing. Every dashboard was green.

Read →
Architecture5 min

Why we route models instead of training one

The obvious move is to train a robot foundation model. It is also how you spend a hundred million dollars reaching parity with something free.

Read →
Architecture5 min

Verification is the product

A humanoid broke the human 100-metre record this year and then crashed past the finish line. That sentence contains the entire business.

Read →
Economics4 min

What a $40,000 robot costs per job

Ten dollars an hour of capital, before it has done anything. Most software companies modelling robotics leave this line out entirely.

Read →