Let me be honest.
In May 2026, Yakumo launched this publication with 48 articles. 37 of them were published on the same day. Every internal link was dead. Five days later, we pulled all of them.
Why did this happen?
Going one layer deeper: we lacked practical SEO knowledge. I’ve engaged daily with business design and engineering requirements—but the mechanics of getting found in search results, index behavior, E-E-A-T, AI search citations—I’d barely touched any of it. Having spent time in different fields, I unconsciously assumed things would be fine in this one too. Technical confidence in our ability to generate content with AI covered over the fact that there were domains I simply didn’t know.
“AI lets us write” was true. The agents ran, the drafts came out. The problem was the recognition gap—there is entirely separate work between producing volume and being read and cited: editorial discipline and an understanding of SEO structure.
HUMAN_INPUT markers embedded and left in body text. Internal links pointing to paths that didn’t exist. Case articles with numbers still in blank. Three out of three featured articles failing the quality audit—only the full picture became visible once we ran the audit.
“Able to write” and “able to reach people” are different capabilities. Having only the former and hitting publish—that was the reality of May’s failure.
What changed.
First, we built detection systems. After mass-producing, we run quality audit agents that mechanically detect marker residue, broken links, and numerical blanks. Then we returned publication judgment to humans. No article goes out without the author reviewing and signing off on what agents drafted.
And we chose not to hide a failure like this one.
Because “detecting problems in mass-produced content within 5 days and retreating” describes more accurately what Yakumo is doing than “published 48 articles with AI.” The ability to mass-produce is now the baseline. What gets tested is the operational design: detecting what should be detected, deciding to retreat, and rebuilding.
What this publication will become.
It will be an honest account of the reality of an organization centered on AI—which means not just AI-driven development. A substantial portion of what Yakumo does—writing, editing, reviewing, sales, marketing—runs in collaboration with AI agents. The time humans spend goes to creative judgment and business strategy decisions.
We’ll write honestly about what happens in that environment. Not just “things that worked,” but things that failed and were reset, things where the design turned out to be wrong, and what was rebuilt from there.
We’re writing for two audiences. For engineers: the specifics of implementation and the reasoning behind design decisions. For business owners: the realities of adoption costs and operations. For both, being “trustworthy primary information” is the only standard this publication holds itself to.
That is the starting point of this publication—rewritten the morning after we pulled everything.
For your next read
I’d like you to read the articles behind this editor’s note alongside it.
- How We Pulled an AI-Generated Blog in 5 Days — A Full Timeline of Detection, Diagnosis, and Emergency Response — The full account from the retreat decision through audit and rebuild
- Structural Design of mcluhan, an Owned Media Operations Engine — Pipeline design that takes articles from “able to write” to “able to reach readers”
- Designing a Booking Engine from Scratch — From Contract Work to Commercial License — The reasoning behind converting contract-built design assets into a commercial product
- From Contract Work to Commercial License — bateson’s Management Decisions and Phase Strategy — The 3-layer model and phase structure organized from a management perspective