Going Public: Six Months of Private Notes, All at Once
Eleven posts went up in one afternoon, each keeping the date it was written on. Why the notes stayed private, why they are not staying private, and what comes next.
Everything below this post was written before anyone could read it. The oldest piece dates to March, the newest to last week, and none of them were published on the day they were finished. They lived in a private repository next to the scripts they describe, which is where notes usually live when you write them for one reader.
Today the whole thing goes up at once. Each post keeps the date it was actually written on, because that is the honest label for it, and I would rather say plainly that the archive arrived in one afternoon than let eleven backdated files imply an audience that did not exist. Nothing here was published earlier. Nothing is pretending to have been.
Why they stayed in the drawer
The useful answer is not modesty. Writing without an audience is a different instrument: you skip the throat-clearing, you keep the parts you would be embarrassed to defend, and you never have to decide whether a section is interesting to anyone but you. That is a good way to think and a bad way to be corrected.
The bad answer is that I did not trust some of the methods yet. A procedure that works on your own properties is not a procedure that works. It is a habit with a small sample behind it, and the difference only shows up when someone with a different site and a different failure mode tells you where it breaks. I sat on the internal-linking material for months for exactly that reason, and on the crawl-scheduling material longer.
At some point sitting on it stops being caution and starts being an excuse. A method you have never had to defend is a method you have not finished testing.
Why they are not staying there
Publishing changes the incentives in a direction I want. Three of these posts commit me to something in public: a pre-registered measurement of AI-answer citations whose data goes up whichever way it comes out, an internal-linking holdout that has to report its null if that is what it produces, and a teardown protocol that promises a named target rather than a hypothetical one. Written privately, all three are intentions. Written here, they are debts.
The other reason is narrower. A large share of what circulates in this field is a confident claim with no instrument attached, and the correction rate is close to zero because nobody can check anything. Almost every post in this archive exists because I could not find a written procedure for something and had to build one. Publishing the procedure is the part that has value; the conclusions I drew from it are the part most likely to be wrong.
Where to start
The archive is not a course and it is not in reading order. Two posts are hubs and the rest hang off them.
If your problem is that pages are not getting indexed, start with the scheduling-versus-selection split. It is the one distinction that determines whether anything else in the indexing cluster applies to you, and most published advice on the topic ignores it. If you have access to your access logs, the crawl pipeline is where the evidence for that distinction actually comes from, and it is the post I would hand to an engineer who thinks SEO is not a technical discipline.
If your problem is that pages are indexed and not ranking, fingerprinting the SERP before you write is the parent of that cluster. The consolidation, decay and internal-linking posts all use its output as an input.
If you have twenty minutes and no appetite for any of this, join your sitemap against your logs. It is the cheapest thing in the archive and it answers three questions that no crawler purchase will answer for you.
What this blog will not do
No case studies with numbers I did not measure. Where a post would be stronger with a chart from a real property, it currently says so or says nothing; it does not invent the chart. That constraint is why several sections here read as method rather than evidence, and it is the constraint I am least willing to relax.
No tool scores treated as targets, no ranking-factor lists, no correlation studies over scraped result sets, and no screenshots of traffic going up as an argument that anything works.
No courses and nothing to buy. If that changes, it will be a post that says so, not a banner nobody announced.
What is next
The measurement window on the AI-citation protocol is open now, which means that post has a deadline attached to it and the results exist or they do not. The first real teardown needs a target chosen and a decision about whether to name it; both are questions of what is fair to the site being examined, not of what makes a better headline.
Beyond that, the schedule is whatever breaks next. Almost everything in this archive started as something going wrong on a property I had access to, and I have no reason to think that supply is drying up.
Corrections are welcome and I would rather have them in public: @g33konaut. If a procedure here fails on your site, that is the interesting case, and it is worth more to me than agreement.