The current moment in software development is pushing many devs toward two paths: microsaas (or a closed product) to monetize quickly, or betting on AI agents and agentic automation — products and tools where the agent "resolves" tasks on its own. This article is about why I didn't go that route. An analysis of the landscape, a conscious choice, and a provocation: the common path is legitimate, but it's not the only one.
What's at stake
There's pressure for revenue, for visible impact, for surfing the wave of the moment. Microsaas has become a shortcut for many indie developers: lean product, subscription, growth. On the other side, AI agents and agentic workflows have gained traction — the idea that AI isn't just an assistant, but an autonomous workforce that reads code, implements features, opens PRs. In 2025–2026 this stopped being just an experiment and became a concrete option for both product and positioning.
None of this is wrong. These are choices. What I want to make clear is: why, in the case of Schepta, I chose a different path.
Foundation before agent, value before price
At some point you've probably seen the same thing I have: devs launching microsaas, others automating everything with agents, the feeling that anyone not on that track is leaving money on the table.
I considered both paths.
- Microsaas brings contracts, pricing, and vendor lock-in before the value is proven. For a problem like this — turning a schema into real UI, stably across multiple frameworks — that's friction before its time. The value needs to be demonstrated first; the business model comes after.
- Agents have a similar appeal: automating interface generation seems like the obvious next step. But agents that generate or modify UI need a stable foundation to operate on — a clear, auditable contract between backend, agent, and frontend. Building the agent before that foundation is fragile. Order matters: first the engine works well, then the agent has something solid to build on.
But a stable foundation has a cost: it needs to be proven. Broken forms and flows impact users and business directly — it's a mission-critical problem. Schepta's 100+ unit and E2E tests exist exactly for this reason: so the engine is reliable enough to be extended, contributed to, and used in production without surprises.
And that was the path I chose. Whether it was the right one, only time will tell.
If this discussion about paths resonated with you, Schepta's documentation and examples are waiting for you at schepta.org.
What about you: are you on the microsaas path, the agents path, or building something else entirely? What drove your decision?
Next step
In the next article in the series we talk about When AI generates the interface: the schema as a guardrail: how the ecosystem (json-render, A2UI) validates this approach, how Schepta responds to the problem of AI non-determinism, and how it fits into workflows ranging from assisted coding to 100% automated agents.
Top comments (0)