The Orchestrator

This is not a prediction that programmers disappear, a prompt-writing manual, or a claim that access to a model qualifies anyone to run a hospital, a water-treatment plant, a nuclear reactor, or a financial market. It is also not a promise that universal basic income arrives tomorrow because a few software companies have discovered a new way to describe abundance.

And it is emphatically not another attempt to make the current sales pitch sound inevitable. The market has supplied enough certainty for several lifetimes. Every few months a new label arrives — copilot, agent, autonomous agent, digital coworker, frontier professional, vibe coder, citizen developer — and a demo becomes a strategy deck before anybody has asked who is going to maintain the thing.

Ed Zitron’s “There Is No AI Revolution” makes the economic objection precisely: weak unit economics, uncertain product-market fit, thin evidence of durable returns.⁷ I think he is right that the narrative has outrun the evidence. What does not follow is that the work is unchanged, and that is the distinction the rest of this book depends on: a market can be wildly unstable while a team’s daily practice has already changed in ways its practitioners are not going to unlearn.

Tim O’Reilly separates human augmentation from the investor fantasy of replacement,⁸ and grounds the architecture of participation in open infrastructure rather than any single vendor’s stack.⁹ Mike Loukides supplies the practitioner’s warning¹⁰ — the hard skill is managing context, not clever prompting — and Cory Doctorow supplies the labor version: the difference between a centaur pursuing a human goal and a reverse centaur pressed into service as a machine’s peripheral.¹¹ Andrew Stellman sorts the name pile with care — vibe coding, loop engineering, agentic engineering are real and distinct practices that need an umbrella, not a winner — and Drew Breunig names the debt every serious team eventually pays: prompt debt, the prototype convenience that hardens into a brittle, model-locked specification language.¹²¹³

Put them side by side and you can see how little they overlap. Zitron attacks the business mythology. O’Reilly argues for human-centered augmentation and the architecture of participation. Loukides presses on context, technical foundations, and maintainability. Doctorow examines ownership, labor, and power. Stellman and Breunig are in the room with their sleeves rolled up, naming practice and debt while the tools change under everyone’s hands. This book takes the whole conversation seriously and then makes its own claim on top of it: the hype may well be inflated or dishonest, and the work is changing anyway, and somebody has to be accountable for the systems doing it.