Nine Journalists Who Never Sleep, and the Editor Who Argues With Them

Nine Journalists Who Never Sleep, and the Editor Who Argues With Them

Ask a CMO what content they wish they had and you get the same answer in three flavors. Something about what happened this week that our buyers will be asked about. Something that teaches our buyers the thing they keep getting wrong. And something with a real point of view, the kind of piece a board member forwards. Now, how, wow. Every marketing team knows the taxonomy. Very few staff it.

That is what we set out to change in the Content Studio this week. Not another "generate ideas" button. A newsroom.

The desk model

A real newsroom is not organized around topics. It is organized around desks, and each desk has a standard. The wire desk lives and dies by the fact spine and the correction. The data desk will kill a story when the sample turns out to be self-selected. The features desk states the mainstream view fairly before it argues with it. The science desk grades evidence and refuses to let a preprint carry a claim alone.

So we built three desks and staffed each with three scribes. Not three prompts with different adjectives. Nine defined roles, each researched against how the real job is done.

1. The Now desk is the newsroom. A wire reporter modeled on Reuters and Bloomberg practice, who ranks sources primary to commentary and will not file a pitch whose so-what needs a hedge. A beat reporter embedded in the practitioner community, with a three-thread rule: a complaint is a pattern only when it shows up in three independent threads across two venues inside ninety days. And a data journalist in the FiveThirtyEight lineage who grades every statistic on provenance, denominator, sample and sponsor, and writes the sentence that says what the number does not show.

2. The How desk is the education desk. A documentation-grade explainer who writes the task statement first and classifies the gap in the top-ranking guide as exactly one of five kinds, none of which is "shorter than ours." An instructional designer who harvests forty verbatim questions per run, clusters them by misconception rather than keyword, and picks a format only after writing the learning goal. And an operator-in-residence who tags every fact by provenance and applies an ownership test: if a competitor could publish the playbook unchanged, it is not ours.

3. The Wow desk is the features desk. A long-form writer in the Wired and Economist register who sorts research, money and practitioner sources into three stacks and looks for the disagreement between them, because convergence is a how-to and divergence is a feature. A science and technology correspondent with a doctorate in measurement science who keeps "demonstrated" and "claimed" in two columns and assigns every innovation a radar ring the evidence can defend. And an essayist who diffs the company's own beliefs against the category's conventional wisdom, builds a named framework with three to five parts, steelmans the opposing view, and names the falsifier.

Beats, not topics

Every scribe works a beat: the search phrases the desk monitors for this specific client. The first time a desk runs for an organization it seeds its own beats from the locked foundation, the market category and curated competitors for Now, the active target queries and persona jobs-to-be-done for How, the persona pains for Wow, and then asks an assignment editor to propose more. The proposals wait for a human to accept them. Nothing is silently assumed.

Each run, the scribe researches its beats on the live web, reads the client's ICP, personas, value propositions, brand voice and signal library, and files five pitches. Every pitch names a persona, the decision on that persona's calendar it changes, the pain, the claim and the gain, and the evidence. The evidence rule is absolute and mechanical: a scribe may cite only URLs that appear in the research dossier it was handed. A remembered URL is a fabrication and voids the pitch.

The editor who argues

Here is the part we care about most. Before a single pitch reaches a human, an editor-in-chief pass argues against all of them, using the filing desk's own checklist. Did the wire reporter bring a primary source. Did the data journalist state the denominator. Is the essayist's steelman a real steelman or a strawman in a good suit. The editor keeps, cuts or merges, scores the survivors, and writes a two-sentence desk note. The reason for every cut is recorded, so dismissals become something each scribe can learn from.

The best pitches land in the Inbox tagged with the scribe's byline. The human votes, dismisses with a reason, or promotes to a draft. When the organization's signal is still thin, the desk holds its pitches instead of queueing them, and the human decides with one click. Autonomy grows with the richness of the foundation, not with our confidence in the model.

Why this is different from a content idea generator

Three things, and none of them is the model.

First, the roles carry standards, and standards are what make output legible. You can look at a wire pitch and ask whether it has a primary source. You can look at a data pitch and ask for the n. A generic "content ideas" prompt gives you nothing to hold it to.

Second, the research is real and bounded. The desks read the live web every week on phrases the client chose, with caching and hard caps, and every claim is traceable to a link the scribe was actually shown.

Third, the loop closes. Every vote and every dismissal reason is captured against the scribe who filed the pitch and distilled into guidance that scribe reads on its next run. The newsroom gets better at this client, not at clients in general.

Discussion items

1. Which of your current content ideas would survive a wire reporter's so-what test: one unhedged sentence naming who has to do what differently on Monday?
2. If a competitor could publish your last three playbooks unchanged, what did you actually own?
3. Where does your team's Wow content state the mainstream view fairly before disagreeing with it?

Questions to ask

1. What are the beats your buyers are actually living through this quarter, and who on your team is monitoring them every week?
2. Which hard question do your personas keep asking that your content keeps answering one level too shallow?
3. What is the one framework your company should own, and what evidence would prove it wrong?