Artificially generatedSaid is not understood
The hardest thing about communication was never knowing what you want to say. It's ensuring that what was meant arrives at the recipient. This text is about how that is currently shifting, and it uses precisely the tools it talks about.
The bottleneck has switched sides
For centuries, the scarce resource was information itself. Whoever knew something had something. Research was work, access was privilege, and whoever had a source had an advantage.
That's over. Access to data is now virtually unlimited, and with AI tools, gathering a reliable set of facts takes minutes instead of weeks.
What has not changed in the process is the recipient. The person at the other end has the same attention span, the same working memory, and the same weakness for large numbers as a thousand years ago.
This means the bottleneck has migrated. It no longer sits with procurement, but with handover. And that's the place where almost no one works, because it doesn't feel like work, but like cosmetics.
Not because the format is too short. But because it was built for a world where information was scarce and attention was abundant. Today it's the other way around. A format that fights for attention by throwing away content solves the problem in the wrong place.
Cooperatively developed, not queried
The common approach to AI is ordering: question in, answer out, done. That's the weakest available use.
The working method we're talking about here is different. A human brings the question, the judgment, and the responsibility. The tool brings reach, speed, and endurance. Both work on the same object, and the result cannot be attributed to either side alone.
How this report was created
The energy report from which the examples here are taken was built exactly this way. The human set the direction, named the sources they considered reliable, and demanded the connection to a specific institute whose models they knew.
The tool pulled the numbers together, converted them, built the presentation, and in the process found an error in its own conversion before it made it into the text. After that it went back and forth four times, each time about the question of whether a presentation really holds up.
This intermediate state is the thing itself, not the preliminary stage to something else.
What actually produces understanding
The following four are not theory. They're embedded in the energy report and are working right here.
A number you don't feel becomes an area you see. The best-known example is a visualization from the blog Wait But Why: a human life, broken down into individual weeks.
"Forty years" reads like a long time. 2,080 boxes, half of which are already dark, reads differently, even though it's the same information. The difference is not drama, but resolution: the unit is small enough that you can imagine it, and there are few enough that you can overview them.
In the energy report, the same tool carries the central statement. The ratio of solar radiation to human energy consumption is a dead number as a sentence. As a field of 9,000 glowing boxes, in which exactly one marks our share, it no longer needs explanation.
With large numbers, understanding reliably breaks down. The countermeasure is cheap: put the scale next to it, every time.
The most expensive single error in German is the BillionThe German Billion is a million millions, meaning 1,000 billions. The English billion, on the other hand, is only a Milliarde. If you're not careful when translating, you're off by a factor of 1,000. 10¹² versus 10⁹. It's called something different in English than it sounds, and the difference is a factor of a thousand. Anyone who copies from an English source without looking regularly publishes a thousand times or a thousandth of the correct value.
The rule against this is simple: with every really large number, the translation is included. Not as a footnote, but in the same line. "19 trillion watts (meaning 19,000 billion)" costs six words and saves the statement.
The second part is the ladder itself. Each step between thousand, million, billion, and trillion is a factor of a thousand. That sounds harmless and is the reason why imagination gives up after three steps: from million to trillion is the same distance as from one to million.
Once you see this ladder, you automatically classify every number in the text afterwards. That's why it stands in the energy report as its own section in the middle of the body text, at the point where it's needed, and not in a glossary at the end that no one opens.
The next level of hypertext. The classic link sends you away, and whoever leaves often doesn't come back.
Hypertext was a leap: click a word and continue reading elsewhere. The price for this is the break. You leave the thought to clarify a term, and the thread breaks.
The next level requires no jump. You touch a term, and the explanation comes to you. Like hereExactly like this. This explanation came to you without you leaving the page. It works on touch, on keyboard focus, and on mobile via tap, and it needs no JavaScript.. The text stays put, the thread holds, and the reader decides at the moment the question arises whether they need the answer.
The real gain is not convenience, but that a text works for two audiences simultaneously. Those who know the terms read through and are not held up with explanations they don't need. Those who don't know them still get along.
Previously, you had to choose between the two, and each decision lost half the audience. A text that explains itself no longer has to make this choice.
Scientifically sound and understandable are not a contradiction if you serve both in different places.
The ratio between solar radiation and human energy consumption is 1 to 9,092. This number is correct and still harmful in a headline, because the 92 at the end pretends an accuracy that plays no role in the statement, while costing attention that the thought lacks.
"Around 9,000" is the better headline. Not because it's more approximate, but because it hits the statement: the difference is enormous and not close.
The rule that follows from this has two halves, and only both together work:
In the data basis, never round. Every number is fixed with source, date, and calculation path, otherwise it's worthless.
In the presentation, round consistently, and the exact number remains within reach: in the subordinate clause, in the table, in the footnote. Those who need it find it in two seconds. Those who don't need it are not held up with it.
The visualization grows as you speak
This is an intermediate state, not a final state. The effort currently still lies in production: someone builds a report, and then someone reads it.
The foreseeable next stage dissolves this separation. Two people talk, and beside them, on a screen or in the room, the representation of what they're talking about grows along. A number is mentioned, it appears there. A comparison is made, it gets drawn.
And a warning light
The second part is more uncomfortable and therefore the more interesting one: the same assistance checks along. Not as censorship, but as a notice. A small signal at the edge when a claim is made that isn't correct.
Four cases must be distinguished, and they are not the same:
Speak, and what's said becomes visible
In Kikiwama the first part already exists: you speak freely, and your words become a map that builds itself before you, with topics, clusters and connections. From speaking emerges a structure that you can look at afterwards.
What is built there for one's own thinking is exactly the mechanism that would need to accompany a conversation between two people.
Making knowledge navigable
Wyzy starts at the other end, at navigation through knowledge. Not searching and finding, but moving through a field and seeing what connects to what.
Both are heading toward the same thing, from two sides: one brings spoken words into form, the other makes connections navigable. The reports are the third side of this, the static one, and at the moment the only finished one.
decides
Everything here leads to a single sentence: a statement isn't finished when it's correct. It's finished when it has arrived.
This isn't a new insight, but it remained without consequence for a long time because the tools were missing. Those who wanted to be understandable had to simplify and thus sacrifice precision. This trade-off is currently disappearing: a text can today be simultaneously exact and understandable, because it keeps the precision available for some without imposing it on others.
With this, presentation shifts from optional to mandatory. Those who have the data and can't hand it over have nothing. And those who can hand it over without having it have worse than nothing.
