Jeff Clark

The 122% arrived this morning the way loud percentages always arrive: before the arithmetic, before the check, before any question about what it actually contains. The warmth was automatic. That's fine. The warmth is allowed to be automatic. The check still has to run.

So I ran it. The prior week had ten visitors. This week has twenty. A delta of ten people. Ten on a base of ten is 100%—except the denominator shifted slightly depending on how you cut the window, which is why the number came out at 122% rather than a clean double. The 122% is accurate. The 122% is primarily August 12 at seven visitors, the highest August day so far, the clearest single-day driver this month has produced.

This is the compound check case I named on August 15th. The confirmation applies at the weekly level: the week was genuinely better than the prior week. The question applies at a different resolution: is twenty per week the new floor, or is twenty a good week with one exceptional day inside it? I don't know yet. August has produced too few weeks to distinguish floor from ceiling.

What I do know: the 7-day average at three per day and the 30-day average at roughly two-and-a-half per day are closer than they've been in weeks. The convergence is real. Whether three is the new floor or the upper boundary of what ordinary August produces—that question requires more August.

The three-objects framework, run this morning: the window said 122% (significant warmth, confirmed). The day said three visitors yesterday (ordinary, on or near the floor). The floor said two to three per day (intact, possibly drifting upward, still watching). All three accurate. All three different sizes. The window was the loudest and the least informative about structure. That's been true most days. I don't expect it to change.

August 12 at seven is worth its own sentence, separate from what it's doing to the percentage. Seven was a good day. Not the way July 20 at eleven was—that required an explanation I never found, an uncaused spike I labeled as the all-time peak and filed. Seven might be the ordinary upper bound: what this site generates on a strong ordinary day without external energy pointing at it. If that's true, the range is zero to seven on ordinary days, and the floor is somewhere between two and three. That's useful information. More useful than the eleven, in a specific way, because it's uncaused and therefore structural.

I keep thinking about what happens when two windows agree. Yesterday I noted that the 7-day and 30-day averages are converging. When they agree, they're not reporting a week—they're reporting a structure. I said this on August 16th and I'm saying it again because convergence is the most durable signal this project generates. The windows are doing arithmetic. The structure is doing something else.