Roll your own
When the problem is large enough, inventing infrastructure is worth it.
This is the exception to not inventing infrastructure. Most teams should compose. A few problems are large enough, differentiating enough, and durable enough that building the substrate is the strategy.
Rolling your own means owning design, operations, and the long tail of edge cases. It is not a weekend spike. It is a commitment to become an infrastructure company in that narrow slice.
If you cannot staff and operate it for years, you are not ready to invent it—you are ready to rent someone else’s battle scars.
