What I learned running a newsroom of one agent and nineteen steps

I built a newspaper with no journalists. Not a demo, not a prototype in a notebook, a real one that publishes on its own, and once earned organic press coverage in two countries without a single human in the newsroom.

Under the hood it is one autonomous agent and nineteen deterministic LLM steps, arranged like an actual editorial hierarchy. The Publisher is the only real agent in the system: it runs a genuine tool-use loop, picks its own tools, decides for itself what gets covered, and decides when it is done. Everything downstream of it is a fixed pipeline. The editor in chief, the section editors, the fact checker, the angle checker and the quality assessor read like a masthead, but each one is a single-shot prompt at a known position in that pipeline. They cannot call tools, they cannot loop, and they cannot decide to go and gather more evidence. It runs on a schedule, defends itself against prompt injection, stays inside a strict budget, and corrects itself when a reader flags a hallucination.

Building it taught me things that no tutorial about agents will tell you, because tutorials stop at the part where one agent answers one question. The interesting problems all live past that point.

The first lesson is that orchestration is the whole game. A single agent doing a single task is easy now. Nineteen steps that have to hand work to each other, agree on what is true, and not contradict each other is a different kind of problem. It looks less like prompting and more like designing an organization. Who decides. Who reviews. What happens when the fact checker and the draft disagree. That last one has a duller answer than it sounds: no checker overrules anybody. Its output feeds a conditional I wrote, and the conditional sends the piece back for a revision or drops it. I spent far more time on the flow between the steps than on any individual prompt.

The second lesson is that autonomy without guardrails is just a fast way to fail. An agent left alone will confidently do the wrong thing at scale. So most of the real engineering was not the clever part, it was the boring, essential part: budget ceilings so it cannot spend itself into the ground, injection defense so a malicious page cannot hijack the pipeline, and a verification loop so a hallucination gets caught and corrected instead of published and forgotten. One agent and nineteen gated steps is that sentence made concrete. Every place I could turn an open-ended decision into a fixed step, I did, because a step can be tested, priced and reasoned about, and an agent can only be watched. The magic is not that it runs on its own. The magic is that it runs on its own safely.

The third lesson is that you cannot improve what you cannot see. With a single prompt you can eyeball the output. With nineteen steps running behind an agent, on a schedule, you are blind unless you build the ability to watch them. Where did the cost go. Which step produced the weak paragraph. Why did this story come out wrong. Observability stopped being a nice to have and became the difference between a system I could actually run and a black box that occasionally embarrassed me.

The fourth lesson is the one I did not expect. Building a fully autonomous system made me appreciate human judgment more, not less. The pipeline is fast and tireless and genuinely good at its pieces. What it lacks is the sense of when something is off in a way the rules did not anticipate. My job, in the end, was to encode as much of that judgment as I could into the structure, and to stay honest about the parts I could not.

People ask if this replaces newsrooms. That is the wrong question, and it misses what the project actually is. It is a working answer to a harder question: how do you get a system built out of language models to do real, consequential work, over time, without a human catching every mistake. That question is going to matter far beyond news, and most of the answers are not in the model. They are in the architecture around it.

That architecture is the work. The newspaper is just where I learned to build it.

Have any thoughts? Send me a message >