The newsroom now takes requests
For most of the last year, the question I got from operators was never the one I expected.
I would explain the newsroom - nine desks, real sources, an editor that fixes what it finds instead of flagging it and walking away - and the person across from me would nod politely and then ask something much smaller.
"I have twenty-five topics sitting in a spreadsheet. Can I just hand them over?"
The first few times, I treated that as a failure to understand the product. It was the opposite. It was a very precise description of the job they actually needed done, and my answer at the time was some version of sort of, eventually, if you set it up right. That is not an answer. That is a sales call.
So this release is about the door - and a door is only worth building for what stands behind it. The nine desks are the product. What changed is how much work you can get to them, and how little ceremony it takes.
What "takes requests" actually means
Mastheads now opens on one field.
Type a line and it is one article. Press Enter for a second line and it becomes a list, one article per row. Delete back to a single line and it is one article again. You never choose a mode from a menu, because a menu is a question you have to answer before you are allowed to start, and the system already has the information - it is looking at what you typed.
There are seven ways in now, and they are all the same pipeline underneath. A topic. A list. A structure you reuse every month, crossed with your own data. A URL you already published. Two products to compare. A subject and a count. A term to explain. One article is one full run of the desks, whichever door it came through.
The one I underestimated is the URL. Point it at something you published two years ago and it comes back refreshed, rewritten, or built to beat it. Point it at your whole sitemap and it lists what you publish, filtered to the pages you have not touched in twelve months. Every operator I know has a hundred posts quietly going stale and no list of which ones. It turns out the most useful thing we could do was not generate anything at all - it was show you what you already have.
The two rules we did not trade away
A door is easy to build badly. Anything that lowers friction can lower it past the point where the thing is still trustworthy. Two rules held, and both of them cost us speed.
Nothing is sent unread. Hand it a list of URLs and every page is fetched and shown to you before a single generation is spent - the headline we found, the domain, the word count, the date. A page we could not read is named and left out of the run rather than written against. That catches three things a URL on its own cannot: a dead link, a paywall, and a section index that was never an article. A sitemap import will happily hand you all three.
We could have skipped the read. It would be faster and it would look identical right up until the moment it produced eight hundred words about an empty page.
A paid-for draft is never thrown away. This one reads backwards, so it is worth being precise. If a desk finds a problem it cannot fix, it retries, and then the article ships with the shortfall written down and shown to you. It does not get silently passed, and it does not get thrown away.
I have argued about this more than anything else in the product. The objection is fair: surely a newsroom that cannot meet its own bar should say so and stop. But I have been on the other side of that. You wait six minutes and get nothing, with no way to tell whether the topic was bad, the sources were thin, or something broke. A flagged draft you can read and fix beats a silent pass. It also beats an empty screen.
Both rules come from the same instinct, which is the only one I really trust in this category: report, never infer. Show what actually happened, including the parts that went badly. The authenticity panel does it, the run list does it - failed runs are in there too, not hidden - and the read step does it before you spend anything.
What this is not
It is not a new product. It is the same newsroom with a much better entrance, and I would rather say that plainly than dress a door up as a reinvention.
It is also not the end of the autonomous side. Mastheads is an autonomous newsroom - that is what the brand has always meant, and the continuous mode is being adapted into this version next. What changed is which door you meet first. On-demand turned out to be the easier thing to understand, and understanding is most of the battle when what you are selling is trust.
If you want to see the read step refuse something, that is the fastest way to understand the whole product. Give it a sitemap. Watch what it declines to write about.
New accounts get ten articles over seven days, free, with no card. That is enough to find the edges.