The minimum viable organization is shrinking.
Historically, ambitious engineering required organizations — because organizations were the only way to aggregate specialized knowledge, labor, capital, infrastructure, and coordination.
Those constraints are changing. Not rhetorically. Mechanically:
- AIcompresses knowledge and labor. Expertise that lived in departments is increasingly callable.
- Commodity hardwarecompresses infrastructure cost. Yesterday’s capital expenditure is today’s used-market purchase.
- Open sourcecompresses implementation cost. Most of any system already exists, maintained by everyone.
- Cloud & distributed systemscompress access to compute. Scale is rentable by the minute — or assembled from what’s already on the desk.
- Modern manufacturingcompresses the distance between prototype and physical object. Boards in days, enclosures overnight.
- Automationcompresses coordination. Pipelines and agents do the meetings.
So the interesting question is no longer simply “What can technology do?”
It is: “What can one person, or a tiny team, do now?”
Nobody knows exactly where that boundary sits. It moves every quarter, and it doesn’t announce itself. The only honest way to locate it is to pick problems that were recently unreasonable for a small team and attempt them for real — with working software, powered-on hardware, and measurements, not decks.
Moonshine Labs exists to continuously test that question. Professional CAD, distributed AI infrastructure, autonomous quality engineering, custom field hardware — these are deliberately different disciplines, because the thesis is about the boundary itself, not any one field.
How work evolves here
EXPERIMENT → PROJECT → PRODUCT
Experiments test a question. Projects are experiments that become substantial ongoing work. Products are the ones that graduate into independently useful commercial or open systems.
Some experiments fail. Some remain weird prototypes forever. Some become open source, some generate consulting engagements, some become businesses. Failure is legitimate output. We are comfortable publishing “we tried this, it didn’t work, and here is exactly where the boundary actually is” — that’s a real result, and it’s what makes the successes credible.
We prove the thesis by building. Everything else on this site is just the notebook.