In a March 2026 essay on nick.cloud, software engineer Nick argues that AI coding agents are rapidly commoditizing the mechanical act of writing code, shifting the real bottleneck upstream into what he calls "the meta" — the invisible layer above the model that encompasses prompting, context engineering, orchestration, evaluation, workflow design, and interface design. His central observation: access to powerful models does not automatically produce useful software. The analogy he reaches for is filmmaking. Owning cameras and editing tools does not make someone a director. The scarce resource is knowing how to combine those tools coherently toward a concrete outcome.
Nick's sharper argument is that the larger opportunity isn't becoming a skilled practitioner of the meta layer — it's productizing it. He envisions domain-specific vertical software that packages <a href="/news/2026-03-14-8-levels-agentic-engineering-framework">all the invisible machinery</a> — context gathering, model selection, task decomposition, output evaluation, revision — so that lawyers, founders, analysts, marketers, and designers can move from intent to usable artifact without ever touching a prompt, a context window, or a routing decision. The user supplies goals and constraints; the product supplies everything else.
That framing has a direct strategic implication Nick spells out: raw model access is hard to defend. Any competitor can call the same API. Products that embed domain knowledge, workflow logic, error handling, and accumulated understanding of what users in a specific field actually need are far harder to replicate. Nick calls this "quiet vertical software" — a pointed contrast to the capability demos that have dominated AI product launches. His bet is that the lasting businesses get built in that quiet layer, not at the model level.