A note on four publication dates in the OperatorBook archive
Four early stories in this archive carried a publication date from before the domain existed. Here is what was wrong, what the dates are now, and the two free lookups that let you check it without taking our word for it.
Updated on August 27, 2026
In this story
Quick answer (2026). A date audit of the OperatorBook archive found four early stories carrying a publication date that fell before operatorbook.dev existed. The domain was registered on June 16, 2026. Those four stories were stamped April 30, May 14, May 28 and June 12, 2026, the earliest of them 47 days before the site it supposedly ran on. The stamps were never real publication dates. They arrived with the initial content import that seeded this magazine, and they have now been corrected to dates between June 17 and June 19, 2026. No story text, quote, figure or source was changed.
What the audit found
OperatorBook publishes founder numbers. We ask people to show us their Stripe dashboards, we print the caveats they insist on, and we say in every piece which figures are self-reported and which are filed. A magazine that does that cannot be careless about its own metadata.
Four of the stories in this archive were carrying a publication date that could not be true. The domain this magazine runs on was registered on June 16, 2026. Any story claiming to have been published before that date was claiming to have been published on a website that did not yet resolve.
The stamp on a story is a claim, and it is the one claim we had never checked.
The four dates were not a rounding problem. The oldest was 47 days early. They also sat on a suspiciously tidy pattern, exactly fourteen days apart, each landing on the hour, running backwards from the week the site launched. That is the shape of a back catalogue assembled in one pass, not of four stories filed over two months.
One of the four gives the game away on its own page. The post-mortem of a failed AI app closes by telling you, in its Sources section, that it is built on recorded interviews conducted in May 2026. It was stamped April 30, 2026. The story said plainly that it was reported after the date it claimed to have been published.
The corrections
Scroll to see more
| Story | Date shown before | Date shown now |
|---|---|---|
| A 90-day post-mortem of a failed AI app launch | April 30, 2026 | June 17, 2026 |
| From freelance to $50K MRR in 18 months | May 14, 2026 | June 18, 2026 |
| Why we killed our SaaS at $12K MRR | May 28, 2026 | June 18, 2026 |
| Marta del Sol on hitting $4K MRR with three AI agents | June 12, 2026 | June 19, 2026 |
Each of the four now carries a short dated publisher's note at the foot of the piece saying the same thing this page says. Beyond that note, the bodies are untouched. The reporting in the 90-day post-mortem and in the $12K MRR shutdown is exactly as it was.
How to check this yourself
This is worth spelling out, because the whole point of a correction notice is that you should not have to take our word for it. Two lookups, both free, neither of them requiring anything from us.
Lookup one, when the domain was registered. Registration data for a .dev domain is published over RDAP, the successor to WHOIS, in the JSON format specified in RFC 9083. Request this URL:
https://pubapi.registry.google/rdap/domain/operatorbook.dev
In the response, read the events array and find the entry whose eventAction is registration. For this domain it reads 2026-06-16T15:24:51.776Z.
Lookup two, what date a story claims. Open any story on this site and view source. Every story page carries an application/ld+json block describing an Article, and inside it a datePublished field. That field is what search engines and language models read as the publication date, and it is documented in Google's Article structured data reference. Compare it against the registration timestamp from the first lookup.
If the second number is earlier than the first, the date is not real. Run it on us again whenever you like. Run it on anyone.
What we can and cannot claim about the new dates
The new dates are the earliest defensible timestamps consistent with the evidence that survives: after the domain existed, after or close to the moment each record was created in our own system, in their original relative order, and before the first story on this site whose date was never in question. They are deliberately spread across real working hours rather than snapped to a tidy interval.
They are not recovered from a publication log, because no such log survives for these four. We are not going to dress an estimate up as a recovered fact. What changed is that the dates are now possible and disclosed, where before they were impossible and silent.
The four stories were the only ones affected. The rest of this archive was checked against the same registration timestamp and came back clean.
Written by
Joaquin del RioJoaquin del Rio covers the money behind the milestones for OperatorBook, digging into what bootstrapped and indie founders actually earn and what it took to get there.
Frequently asked questions
Which OperatorBook stories had their publication date corrected?
Four: 'A 90-day post-mortem of a failed AI app launch' (April 30, 2026 to June 17, 2026), 'From freelance to $50K MRR in 18 months' (May 14, 2026 to June 18, 2026), 'Why we killed our SaaS at $12K MRR' (May 28, 2026 to June 18, 2026) and 'Marta del Sol on hitting $4K MRR with three AI agents' (June 12, 2026 to June 19, 2026). Each carries a dated publisher's note explaining the change.
Why were the original dates impossible?
The domain operatorbook.dev was registered on June 16, 2026, at 15:24:51 UTC, a fact published over RDAP by the .dev registry. All four stories were stamped earlier than that, the oldest by 47 days, so they claimed to have been published on a website that did not yet exist.
Was any of the reporting changed?
No. The only edit to each of the four stories is a short dated publisher's note appended at the foot of the piece. Every quote, revenue figure, caveat and source line is unchanged, including the notes on which figures are self-reported and were not independently audited.
How can a reader verify the corrected dates independently?
Two lookups. Request https://pubapi.registry.google/rdap/domain/operatorbook.dev and read the events entry whose eventAction is registration. Then open any story, view source, and read the datePublished field in the Article ld+json block. If datePublished is earlier than the registration timestamp, the date cannot be genuine.
Where did the wrong dates come from?
They arrived with the initial content import that seeded this magazine two days after the domain was registered. The four dates sat on an even fourteen-day spacing and on round hours running backwards from launch, which is the signature of a back catalogue built in a single pass rather than of stories filed over several weeks.
More stories
A 90-day post-mortem of a failed AI app launch
The app worked. The demo hit 140,000 views and 412 people signed up on launch day. Ninety days later: 11 paying users, ~$209 MRR, and an $8,000 monthly burn. This is the boring middle where AI products actually die: the retention cliff the launch hid, the month spent building the wrong thing, the five user calls that came too late, and the specific trap of an AI demo that's too good to be true.
Why we killed our SaaS at $12K MRR (a post-mortem)
Cadence reached $12,400 in MRR with 140 accounts and an up-and-to-the-right graph, then the founders shut it down on purpose. This is the post-mortem of the most dangerous number in startups: too much to walk away from, too little to live on. The retention they didn't track, the customer they optimized for and shouldn't have, the fork they took too late, and the unusually honest way they ended it.
Marta del Sol on hitting $4K MRR with three AI agents
Marta del Sol runs a one-person operations studio from Valencia and crossed $4,120 in MRR across nine clients, with three AI agents doing the delivery and a $612 monthly software bill. Here's the real arc: the underpricing she's embarrassed by, the month two clients churned at once, why her clients renew for the Friday report and not the robot, and the caveats she insisted we print next to every number.