Why no document can be written for every reader you have never met, and what changes when it can be asked instead.
You land on somebody's website. Within a few seconds you decide whether this person is worth any more of your time, and you make that decision on almost nothing. All you have is a headline, a photograph, the first line of a paragraph you have not read yet.
That is not a failure of attention. It is the only decision available.
"You cannot know whether you want to know somebody until you already know something about them, and knowing something about them costs attention you have not yet decided to spend."
The first judgement is usually made too early and on too little, and what it is actually made on is a guess about whether any further effort would be worth it.
I have spent a long time on the other side of the same problem. Writing about what I do means deciding, in advance, what to leave out on behalf of a reader I have never met and cannot ask. Everything I might want somebody to know exists. It is in the projects, in the interviews I have already sat, in the courses, and in the answers I would give if anybody asked me. None of it is missing. What is missing is somewhere to put it.
The top and the bottom have to meet in the middle, on the terms that the middle is willing to have them meet.
A CV has one price and you have to pay it before you know whether you want to. That is why writing it better does not solve the problem. A perfectly written two pages still asks for the whole decision up front, from somebody who has no basis for making it yet.
And because the document is finished before any reader arrives, it has to serve all possible readers in exactly the same way. The person who cares about how a project is run and the person who cares about how a system is measured are both reading the same two pages, and each of them is mostly reading the other one's half. There is no length that escapes this. Make it shorter and it stops meaning anything. Make it longer and nobody reaches the end. Every version is a compromise struck between readers who will never meet each other.
I used to think this was a problem with me: that I had not yet found the right way to say it. It is not. It is a property of the format we inherited.
None of this apparatus is old. Before any of it, if you wanted to know whether somebody could help you, you went and met them and asked. You started with whatever you happened to care about, they answered, and you asked the next thing. Nobody handed anybody a summary of themselves at the door.
We stopped doing that because it did not scale. One person can only have so many conversations, and a document could be written once and read by anybody. That was the correct trade at the time, and the whole structure we now take for granted follows from it: the pages, the sections, the search results that deliver you to a paragraph somebody wrote years ago for somebody else.
What is easy to lose is that the document was the substitute. It is not the natural form of getting to know what somebody does. It is what we adopted when the natural form was unavailable, and we have lived with it long enough that it has stopped looking like a workaround.
The reader's half of this is obvious once it is said: you cannot ask a page anything. If a sentence is unclear, or you want to know why something was done that way, there is no mechanism. You can read it again more carefully, and that is all.
The writer's half is less obvious and I think it is worse. I have no idea what anybody reads. I do not know which paragraphs were reached, which ones landed, which ones were skimmed past, or whether the page was read at all before somebody left. I know nothing about the one thing I actually care about.
What comes back instead is a set of numbers. Impressions. Clicks. Time on page. They are captured, sorted, and available to me on demand, and not one of them can tell me whether a single person understood anything. I have recently spent yet another afternoon looking at those numbers. They are not measuring what I want to know. They are measuring what happens to be easy to count.
This is the fault I enjoy finding in other people's systems. A measurement that was reasonable when it was set, that has been running long enough to look authoritative, and that answers a question adjacent to the one being asked. It is unsettling to find it sitting under your own website, but it is the same fault, and noticing it is not optional once you have made a career out of noticing it elsewhere.
A conversation removes both halves at once, and it does it with one mechanism rather than two.
The reader stops paying up front and starts paying in instalments. You ask one thing. You get an answer worth having, or you do not. Then you decide whether to ask a second. Nobody has to commit to the whole record in advance, and nobody has to read somebody else's half to reach their own.
That is the general case. Here is the version I actually built. The same exchange comes back to me. I am sent a summary of what was discussed and the full transcript underneath it, so I can join at speed and go back through it properly when I have the time. I am not guessing at what somebody wanted. I am reading what they actually asked, in their own words, and picking the conversation up where they left it.
That second part is not a feature added alongside the first. It falls out of it. The thing that lets the reader steer is the thing that tells me where they steered, and it is the same object.
The summarising does not stop happening. A conversation compresses too: nobody gets the whole record, and nobody wants it. What changes is who is doing the compressing and whether anybody can see it. It is done from what I actually wrote, in answer to the question somebody actually asked, and I get to read both afterwards.
The assistant on my contact page, Ada, is where this currently sits. Two things about that are worth being plain about. Ada is not me and does not pretend to be, and a conversation with Ada is not an interview with me: it is a way of asking the record questions instead of reading a summary of it. And this is not a support chatbot in a different jacket. A chatbot is a better index into a fixed set of pages, which is a useful thing and a different thing. This is not indexing the pages. It is an attempt to make the pages less necessary.
Now you know something about them, and none of it cost attention you had not already decided to spend.
I am not going to pretend I have arrived somewhere. What I have built is early, it is unfinished, and the record it draws on will never be complete, because my work is not finished either.
What I hope is further down this road is a site that is mostly not a site: a landing and an interaction, and then a conversation with me. Something closer to the thing the document replaced, reached by going forward rather than by going back.
I am aware that I am still saying this on a website with pages on it, and that the pages are doing most of the work today. That is the honest state of it rather than something to hide. The distance between where this is and where I think it could go is not a failure. It is the part I find worth doing.
If any of this is the kind of thing you would rather ask about than read about, that option is on the contact page.
Start a ConversationI talked this page through with an AI rather than drafting it, which is not far off what the piece argues for. The caption under the first figure is a sentence I said in that conversation, before either of us realised it was a caption.