The operating standard
Structural engineers don't hope a bridge holds. They specify the load and over-build what fails first. This is my version of that document, applied to running software companies.
01DESIGN LOADS
Seven rules I don't renegotiate
Customers are the only investors
Every euro comes attached to a customer I have to serve. That's the right incentive structure, and the only one I use. When revenue is the only capital, customers are impossible to deprioritize.
Nothing is for sale
An exit isn't a milestone. It's the moment a company's interests and its customers' interests split. Remove the exit and a whole category of bad decisions never gets proposed.
Overbuild the boring parts
Backups, billing, uptime, documentation, support response times. None of it demos well. All of it decides whether customers are still around in year ten. I overspend on the parts nobody screenshots.
Complexity is corrosion
Every microservice, vendor, and dashboard added today is maintenance debt compounding quietly. Complexity is rust: inevitable, manageable, fatal if ignored. My default answer to new components: no.
Collect nothing you'd be embarrassed to explain
The test: could I explain every byte I store to the person it describes, without flinching? Products pass by design, not by policy. Data I don't have can't leak, can't be subpoenaed, and can't quietly become the business model.
Slow is a strategy, not a failure
Compounding modestly for twenty years beats spectacular growth that dies in year four. I plan in decades, review in quarters, and don't let the second number bully the first.
Stay small enough to care
Headcount isn't a scoreboard. A small team that knows its customers by name beats a big one reading dashboards about them. I hire when the work demands it, not to fill an org chart.
02STRESS TEST
The questions I ask before every decision
Most bad decisions in software aren't stupid. They're reasonable under the wrong time horizon. A dark pattern is rational if you're optimizing a quarter. A data grab is rational if you're dressing up for acquirers. An unmaintainable feature is rational if someone else inherits it.
My checklist forces every idea through the long horizon first. Anything that only makes sense if I sell, quit, or stop caring fails at intake.
Five questions. They kill about half of what I come up with. The half that survives is the company.