15 years of building the same wrappers.
For 15 years we built custom software for founders, mid-market companies, and the occasional enterprise. The pattern repeated: clients arrive paying $50K–$500K/yr for SaaS that, underneath, was a UI on top of an API they could call themselves. Klaviyo on top of AWS SES. Datadog on top of open-source LGTM. Harvey on top of GPT-4.
The wrapper margin used to be defensible — SaaS vendors had genuine integration value, support polish, ecosystem effects. In 2024–2026, that defense started cracking. AI tooling let small teams ship custom replacements in weeks instead of months. The infrastructure (Postgres, AWS, Anthropic Claude, Twilio, Postmark) became commoditized. The "build vs buy" math flipped in build's favor for any tool with $20K+/yr spend.
Upstream is the productized version of what we'd been doing one-off: rebuild specific SaaS layers on direct billing, fixed-price, fixed-timeline. The catalog of 24 single-tool rebuilds + 5 combo bundles came from looking at our client base and identifying the wrappers we'd already replaced multiple times. Each rebuild is a pattern we've shipped before.
The AI-paired senior engineer model is the productivity dividend that lets us quote fixed prices below where hourly billing would land. We don't pass savings to ourselves — we pass them to clients who get the same code at lower cost than agency-rate would imply. Full method →