EXPERIMENT 002 / PHYSICAL COMPUTE
MICROBLADE.
I wanted a cluster I could physically build, expand, and understand. Small computers, added one at a time, inside a rack I could design and print myself.
SYSTEM PROFILE / 01
A cluster that can grow without becoming a mess.
THE IDEA / 02
Start with the physical problem.
The first version of MicroBlade sounded simple: mount a few compact computers in a modular rack and keep adding nodes later. Then the real work showed up: screw placement, clearances, airflow, cable routing, print orientation, and a build plate that was always just a little smaller than I wanted.
That turned the rack into part of the experiment. If the physical system is predictable, networking and orchestration get easier to reason about. If it is not, every new machine creates another little infrastructure problem.
WHAT I AM LEARNING / 03
The hardware keeps answering first.
Can I add one machine without redesigning everything?
That is the core test. New nodes should slide into the system without turning wiring, cooling, or mounting into another project.
What failed first?
The physical details. Fit, print orientation, warped parts, screw geometry, cable paths, and printer limits all showed up before cluster software became the interesting problem.
Why build the rack at all?
Because infrastructure is easier to experiment with when it is tidy, repeatable, and serviceable. The rack is meant to make the computers easier to use, not merely nicer to look at.
WHERE IT IS NOW / 04
The design is finally becoming repeatable.
The rack is now moving through real prints and fit tests. Each revision has made the design less theoretical: parts have warped, supports have failed, clearances have been adjusted, and the mounting system is getting simpler because of it.
NEXT NODE / EXPERIMENT 003