What are you actually building?
A version of this post appears on the Web Frolic blog.
The short answer: a ground-up identity provider.
The more honest answer: I keep coming back to Conway's Law, the idea that systems end up mirroring the communication structure of the organizations that build them.
Seams in the product
Most identity products in use today were built by many different teams over many years. You can see it in the product: seams where different groups bolted on features, architectural inconsistency from one part of the product to the next, and real business needs that never became first-class citizens in the design. They got retrofitted instead.
Two examples that stuck with me
- Service accounts have always mattered, but they were always treated as secondary to human users. They're mandatory in break-glass scenarios and show up everywhere in integrations, so people just shared credentials for them. The rise of AI agents made that gap impossible to ignore.
- Test users. Most platforms model a user's lifecycle carefully but miss the lifecycle of a test user. That's not a minor omission. It's foundational to rolling out any product with real confidence.
These aren't edge cases. They're symptoms of identity products shaped by org charts instead of by the actual shape of the problem.
What's next
I'm still heads-down and building, with more to share soon. If you've run into your own version of these seams, I'd love to compare notes.