AI Has Surpassed Humans at Analysis. Even So, the First Piece Decides It

Have AI analyze what you just made — ask it "so, how did this land?" — and it hands back startlingly sharp insight. From a wall of numbers, it separates the lines that are rising from the lines that are sinking, and calls it: "there's a bud here." Let me be honest. For that post-work analysis alone, AI has already surpassed humans. As a human, I'll admit it plainly.

But the more I leaned on it, the more I noticed the opposite. However sharp the analysis, if the ground you put the work on is wrong to begin with, nothing can be drawn out of it. Analysis won't carry you past the edge of the material itself.

So the last thing left to the human hand in the AI era is — the design of "how you put it out" (your output method) of what you made: where you put the first piece out, for whose benefit, and with what kind of fun. Today I write about that — with this site's (passed.jp's) honestly rather embarrassing early numbers, and with the very rules for "how you put it out" that I used to write this article, opened up as before → after. Let me place the conclusion up front. The weakness of your start is not your failure. Don't break — pour your strength into the design of the first piece.

AI's analysis has already surpassed humans (in the post-work)

After you put something out, have AI analyze the response data, and it picks up, in seconds, correlations a human wouldn't notice after staring for half a day. "People who came on this word drop off here." "This one piece, and only this one, clearly holds attention longer." Give it the material, and the analysis of how you were read is fast, and accurate.

Honestly, on this one point, I don't feel I can beat AI. I admit it. Reading the meaning of what happened after you moved your hands is now AI's domain. Without warping it with emotion, coldly, it sorts the lines out of the numbers. This is no longer a place where humans compete.

This isn't a bitter story. It's good news, if anything. It means you can hand the whole heavy labor of analysis over to AI. And if so, the place where humans should keep their strength is somewhere other than analysis. Where that is — this article is going to track it down.

But the more I leaned on it, the more I saw — if the ground is wrong, analysis does nothing

What I learned from heavy use is that AI's analysis has a clear limit. Analysis can only look within what already exists.

Take a piece you put out in the wrong place, have it analyzed all you like, and what comes back is only inside talk: "this heading is weak," "people drop off on this word." It won't yet come out and say, "the real mistake was putting this on this ground at all." Fix the title, change the opening — post-work optimization moves the needle only when the original ground is right.

On the wrong ground, any fine-tuning just changes the speed of sinking a little. However much you polish the post-work, you can't take back a mistake in how you put it out in the pre-work. That was the wall that came into view the more I leaned in.

What you newly put out will, at first, almost always fall silent

Here, let me lay this site's real numbers bare, honestly. It isn't a flattering picture, but I'll write it as it is.

About half a year. Clicks that came from search: four, total. Times it was shown: about 1,100. As a click rate, 0.4%. And — nearly half of those impressions (about 45%) were an accidental exposure on a single English word, "passed." That's just showing up to someone looking up the past tense of "pass"; it isn't someone who wants to read this site. Strip the noise, and the people who actually came and read are a tiny handful.

I think this looks like failure. But it isn't. What you newly put out has, at first, no authority, no links from other sites, no accumulated track record — nothing. So at first, neither impressions nor rank come. This isn't failure; it's the standard form of a start. Whoever begins, everyone passes through here first.

And even within the silence, there is a line. Impressions rose steadily. 145 in early summer, 401 in high summer, 501 in early autumn — more than tripled in half a year. The numbers are small, but the area you occupy is genuinely widening. Silence is not nothing.

Even so, you can draw out where the upside lies, from the start

So are weak early numbers meaningless? The opposite. Weak as they are, what you can draw from them is large. Three things to do.

(a) Narrow, from the very start, to ground you can win

The only piece on this site that took real clicks was the first-hand account of making things together with AI. A true story only I could tell, that no one can copy. Conversely, on ground everyone writes — exam-day know-how and the like — it sank cleanly to rank 40, rank 90.

The lesson is cruelly clear. On ground someone else could also write, the latecomer can't win. The bud is only in first-hand experience that no one else holds. First-hand experience is what you actually did, stumbled on, and verified yourself. This alone can't be copied — not by AI, not by those who came first. Ground you can win means, in short, "the place only you can write."

(b) Have AI analyze the weak data

Here is where AI works. "Impressions are climbing, but clicks aren't. Why?" "Which words are real readers, and which are noise?" The weaker the data, the more humans read it warped by emotion, ending it with a single "that didn't work."

But AI, coldly and accurately, points to where the upside lies, faster and more precisely than a human: "the thing the one click-earning piece has in common is this," "this 45% is noise, so drop it out of the denominator when you read." This is the practice of that earlier "in the post-work, AI has surpassed us." Just prepare the raw early data, and you can hand the whole analysis to AI. The human takes the result and decides "where to put it out next."

(c) The area only grows with time × persistence

One important distinction. "There's little you can do right now with post-work optimization" and "you can't win the area in the future" are two different things.

Authority, and links from other sites, can't be stacked in a day. Put out one piece, then another, and — the way impressions tripled month over month — the area only widens with time and persistence. AI makes analysis fast. But it can't make time itself fast. This, and only this, remains as a domain where the human simply has to stack it, patiently.

Actually, "how I put out" this very article was designed too ―― let me open the rules I used

Up to here I've written "design the first piece" abstractly. But where, in the end, does design actually show up? The answer is almost anticlimactically concrete. In the wording, one phrase at a time. A single word in the title, the verb in a heading, the framing of a fact — design lives there.

I'll confess: this article had a "first design" too. Before I started writing, I made a small table of rules — what to make the subject, in what wording, in what notation, and to whom I'd write. Let me open that table as is. The very text you are reading was built by keeping these rules. I'll lay each out as "a rule (a line you can take home as is) + how I applied it in this article (before → after) + why it works."

Rule 1: Pull the title and the framing word toward the subject. Decide, in a single word, what this is about — before you write.

before (framing): "How to analyze with AI"
after (framing): "the design of how you put out (the output method) what you made"
why it works: when the framing word wavers, the whole article scatters. Fixing the subject to "how you put it out" bundled everything — the analysis, the numbers, the wording — onto the single point of "how do you put it out." The framing word is the name of the article's spine. And the reader, too, grasps in a single word: "this is about me."

Rule 2: Align the wording with the aim. Same meaning, but choose the word that fits the point.

before: "analyze with AI"
after: "have AI analyze"
why it works: "with AI" is a way of speaking where you hold AI as a tool; "have AI analyze" hands the subject over and entrusts it. The spine of this article is "the post-work analysis, you can hand it whole to AI." That point carries through in a single change of grammar. So it held to "have it analyze" from start to finish. Consistency, too, is part of the design.

Rule 3: Unify the notation to one. Don't create spelling drift.

before: "first-hand," "firsthand," and "first hand" all mixed in
after: one spelling, held fixed
why it works: spelling drift creates, in a place unrelated to the content, a faint snag for the reader every time. A snag is a seed of leaving. Unifying puts the reader in a state where they can focus on meaning alone. Design reaches even into a detail like this.

Rule 4: Avoid words of overstatement and hype. Don't bait with "guaranteed" or "anyone, 100%."

before: "This will definitely make your start take off!"
after: "The start is weak even with design. But it becomes a reason not to break."
why it works: hype may take a click for a moment, but it shaves the trust of the reader who finishes. The lead role on this site is honesty. A piece that baited with exaggeration borrows against the trust of the next piece. Push flatly, with the strength of the fact.

Rule 5: Match the tone to the reader. Decide, clearly, on one addressee.

before: (a lecturing tone) "one should do X"
after: (the tone of someone testing it right beside you) "here's what I did. I think it'll work for you too."
why it works: the addressee of this article is "an individual who has just begun putting something out." Not a teacher, but someone standing at the same start, I write as one of the party. That's why I don't hide the embarrassing numbers. Only the words of someone standing in the same place reach the person about to break.

Rule 6: Don't judge good or bad by the weakness of the start. Don't read a stretch of small numbers as "failure."

before: "Four clicks in half a year. It's hopeless."
after: "Four clicks in half a year. This isn't failure; it's the standard form of a start. The area grows with time."
why it works: the fact is the same — four clicks. But the former breaks you, and the latter becomes grounds to keep going. The start is weak for everyone. Pass judgment there, and you throw away the winning ground along with it. Always put the variable of time into your evaluation. But — wording that keeps you from breaking stands only on top of an accurate fact. Don't fudge the numbers; change only the frame. Cheering with a lie isn't design; it's window dressing.

None of these six takes special talent. What it takes is to stop, once, before you put it out, and ask: "this one word — is it standing on ground I can win, or ground I can't?" Design, before any grand strategy, was this stack of small pauses. And — you may have noticed. This article was written by keeping the six I just listed. The very text you're reading is an example of "the first design." Whether it worked out, I'll find out from here. But the one point — that I designed it before putting it out — is certain.

So pour your strength into "the design of the first piece"

Fold everything so far into a single point. If AI has surpassed humans at post-work analysis, then the place humans should pour strength into is the pre-work. How you design and put out the first piece. The human hand remains here, whole. And those six rules, in the end, serve a multiplication of three.

Ground you can win — narrow to first-hand experience only you can write (Rules 1, 2, 6). Miss this, and the other two don't work.
The reader's gain — what's left in the reader's hands when they finish (Rules 3, 5). A piece with no gain doesn't even get read, let alone analyzed.
Fun — while reading, do they want to go to the next heading (the honesty of Rule 4, and the clarity of the framing)? Fun is the one force that stops people from leaving.

Grasp these three first, then put it out. Only then, when you have AI analyze the post-work, does the analysis have material it can actually work on. The order is this: design (human) → put it out → analyze (AI) → design again (human). The starting point of that rotation is always the human's design.

There is no reverse. Take a piece that missed the ground, have AI analyze it all you like, and what comes back is only an accurate death certificate: "this won't grow." Miss how you put out the first piece, and AI's analytical power is wasted, whole.

Even if you can't win, to keep taking it on — on ground you can win — is what grows you

One last honest thing. I've written "design the first piece" up to here — but even with design, the start is still weak. Remember: four clicks in half a year. Design does not make the start strong.

Then what is it for? Design gives you a reason not to break when the start is weak. "This isn't failure; it's the standard form of a start." "The ground I chose isn't wrong — the rest is a matter of time." It becomes the grounds on which you can say that flatly (yes — it means turning the reframing I did in Rule 6 onto yourself). What divides the one who breaks from the one who keeps going is not talent, I think, but this single point.

And this is the same thing I've kept writing on this site. Even if you can't win, the time you spent taking it on — on ground you can win — doesn't vanish. It becomes flesh and blood. The half-year in which impressions tripled, small as the numbers were, genuinely widened the area. The time you tried, even if it tips toward the side later laughed at as "those numbers were tiny," doesn't vanish.

AI's edge in analysis is real. I admit it. But whether that analysis becomes something that works or something wasted comes down to where you place the first piece — down to the human's design. It's fine to lose to AI at analysis. The one point you can't afford to lose is this: choose ground you can win, don't break, and put it out again. That alone is something only you — not AI — can do.

— Honestly, this article too, I'm putting out without knowing whether it'll win. But the ground I chose, I chose to win on. The wording, too, I designed one word at a time. The rest is just to put it out again.