The landing at senku.im states the claim in one line: SaaS is dying, and capital corrupted it. This page carries the argument behind it, four mechanisms drawn one by one. It is a worldview from one builder's seat, after fifteen years inside SaaS. No industry data is measured here; the exhibit is one household's running system, linked at the end.
Software for strangers cannot feel its own flaws. Need reaches it through interviews, surveys and a product manager's reading of them, and whichever interviewee that manager likes best speaks loudest. The reading becomes the roadmap. An owner-built system closes the loop in one seat: the person who feels the flaw is the person who edits, the same hour.
A requirement in a company that sells to strangers is a chain of translations. Users are studied: interviews, surveys, session recordings. A product manager selects which of those voices matter, and the selection is where bias enters. A manager who enjoys talking to one representative of the target base ships that voice as everyone's product, and the product gets ugly fast. By the time anyone who writes code touches the requirement, it has passed through design and planning too. Every hop is honest work. Every hop also moves the product one translation further from anything a user felt.
The chart with designated roles puts decision at the top: designer, product manager, engineering manager. Below that layer sit the people who execute. The executor reads the ask, sees where it is wrong, and builds it anyway, because the structure routes asks down and routes nothing back up. I spent fifteen years on the executing side of that chart; the claim about where the discomfort lives is lived, and it is still one person's account.
Nobody in the chain is careless, and that is the point. The designer, the manager and the engineers each do the job the structure defines. The structure itself decides that the people closest to the flaw hold the least authority over it, so the flaw ships. A system built by its owner deletes the structure: authority and sensation sit in the same chair.
There is a second cost inside that decision layer, before anything reaches a builder. Each role arrives with its own obsession: the designer with one aspect, the product manager with another, the engineering manager with a third, and none of the three has to be the thing a consumer cares about. What the roles produce together is a negotiation, and the negotiation is the product: a compromise on the good days, an ego contest on the normal ones, because that is how the corporate game scores itself. The famous techniques of the software giants get adopted by companies a fraction of their size, and adopted poorly, so the ceremony survives while the judgment it was built to carry does not. An owner-built system has nothing to negotiate: one obsession, the owner's, checked against use the same day.
A subscription prices adoption, and the invoice is the visible cost. The unpriced one is learning the vendor's design decisions, decisions tuned to an average of thousands of strangers, and that learning never completes because every redesign resets it. Purpose-building carries its learning inside the work: by the time the system runs you already know it, because you decided it. The knowledge arrives as a side effect, at no extra cost.
Custom code has a reputation for being hard, and the fear is the learning curve. The curve exists on both routes. A bought product is somebody else's system, learned from the outside, relearned after every redesign, forever. A purpose-built one is climbed once, during the build, and the climb produces the system itself. The vendor's product also cannot stop being generic: its design decisions answer to an average of thousands of strangers, so for any single owner they land wrong in a hundred small ways that no settings page can fix.
The claim has a boundary. Low-level services survive where scale and compute are real: this site runs on rented compute, and its storage is rented too. What is dying is the product layer for strangers, the layer whose hardest engineering is keeping thousands of tenants separated, billed and safe from each other. Built for one owner, that layer has no reason to exist; the landing carries the sketch of exactly what evaporates. And the argument itself is a worldview: nothing on this page measures the industry, it argues from one seat and says so.
The evidence is one family's operations, live since 2026-08-03: a work board ranked by my verbatim words, with a build that fails when they are missing; rules written as tests, each shown to catch its own violation; agents that write the log line before the step, so a crash marks the exact resume point. The floor is at mission.senku.im, the story behind it with its own diagrams at the mission story, the public board at the live board.
An argument page. The mechanisms are one builder's reading of fifteen years in the industry; the exhibit behind the links is real, one household big, and young. Nothing here has a second household behind it, and the pages linked above list their own open questions.