I never learned HTML.I don't plan to.

For months I used AI the way most people do. Summarize this. Write that. Give me the bullet points. It worked every time, and the output was always the same shape. A block of text I half read and forgot. I would scroll, nod, and close the tab. A week later I could not have told you what it said.

Then I changed the ask. Instead of a summary, I asked for an HTML file.

What came back was not a document. It was a thing I could use. The numbers sat in a table I could sort. The risky parts were red and the safe parts green, so I could tell them apart without reading every line. Each section had a tab, so I was not scrolling to find one thing. There was a control at the top that let me change one input and watch the rest move.

I sat there for a second. Then I deleted the prompt I had leaned on for a year and never went back to it.

I never learned HTML, and I don't plan to

That is not a confession. It is the point.

I am not an engineer. Last month I made more than a hundred HTML reports, and I did not write a single line of the markup in any of them. Dashboards, briefs, cheatsheets, standup pages, calculators. All of it came from describing what I wanted and asking for an HTML file.

The skill was never the syntax. The skill is naming the artifact. Make an HTML dashboard for these ad numbers. Turn this into a one page brief with the risks in red. Build a calculator where I can change the price and see the margin. The model writes the code. I decide what the thing should be.

An engineer reading this is already typing the reply. The HTML underneath is not clean. The structure is whatever the model felt like that day, and if you opened the source you would wince.

I know. I do not care, and here is why. Most of what I make matters for about a week. A dashboard for one standup. A brief for one decision. Nobody is going to maintain it, extend it, or read its source. It has one job, which is to be clear to a person for a short time. Generated HTML is good enough at that job that the mess never reaches the screen. If I were shipping a product to a million people, I would want an engineer and clean code. I am not. I am sharing a page.

A week of it

Let me show you a normal week, because the abstract version undersells it.

Monday, a pile of ad numbers from three channels. The old me pasted them into a doc and wrote a paragraph nobody reads. Instead I asked for an HTML dashboard. Cards for spend, installs, and cost per install, with the channels over target sitting in red at the top. In the standup we talked about the numbers instead of hunting for them.

Later that week I had to price a plan. Pricing is where a static table lies to you, because the whole question is what happens when you move one number. So I asked for a calculator. I dragged price up, pulled conversion down, watched revenue follow, and found the soft spot in about a minute. That used to be a spreadsheet and an afternoon.

The requirements doc was the one that surprised me. In plain text it is where decisions go to die. As HTML I gave it tabs for problem, scope, and open questions, and I put everything we had not decided yet in yellow. Two people commented on exactly the yellow parts. That is the entire reason I marked them.

Same person, same week, no new technical skill. The only thing that changed was the format I asked for.

Why it lands harder for someone like me

A wall of text makes you hold the structure in your head. You search it, scroll back to check a number, lose your place, start over. You are the one keeping it upright.

An HTML version puts the structure on the screen. Sections become tabs. A comparison becomes a chart instead of a paragraph about a chart. Severity becomes color instead of a word you have to catch. The first time a teammate opened one of my files and went straight to the red items without reading the rest, I understood what I had actually been making. Not a nicer document. A faster way to look.

An engineer can hold a dense doc in their head. I cannot, and I do not want to spend my attention there. The format does that part for me, which leaves me free to think about the decision instead of the layout.

It just opens

This is the part I underrated for too long.

An HTML file opens. Anywhere, for anyone, with nothing installed. I send one to a teammate, they double click, it is there. No account. No app. No question about which version they have. It works on the engineer's machine and on my mom's laptop in exactly the same way.

It is also hard to break. Miss a tag and the page still renders, usually fine. A mistake degrades into a plain block instead of a crash. The few times I tried to write code myself, one missing bracket took the whole thing down and ate an hour of my night. HTML does not do that to me. It bends instead of snapping.

After years of spreadsheets that look different on every screen and decks that break the moment you open them somewhere new, a file that just opens felt close to cheating.

The AI makes it. I decide if it is any good.

I should be clear about what this is not. The model has no taste. It will hand you a clean, confident layout that says the wrong thing without blinking.

The files my team reacted to were never the ones I made fastest. They were the ones I went back and shaped. I cut sections that did not earn their place. I fixed the framing. I told it the chart was misleading and asked again. The first version is a draft, and a draft is wrong more often than it looks, because a tidy layout hides weak thinking better than a messy paragraph ever could.

What caught me off guard was who noticed. I expected the non-engineers to like this. I did not expect the engineers to be the ones asking how I made them.

We build uselink the same way

Here is the part that still makes me smile.

uselink is a place to host and share HTML, so other people can open it and comment without signing up. We are building a home for HTML artifacts. And we plan the work, run the numbers, and write the specs for it using the same HTML artifacts. The dashboards that track the launch, the briefs that settle a feature, the standup pages, all made the way I just described.

The tool and the way we work on it are the same shape. It is the loop you see when a team builds a coding agent and then uses that agent to build the next version of itself. At some point you trust the thing enough to point it at its own work.

Building this way is also how we found the gap the product fills. I kept making these files and then could not share them. I attached HTML in Slack and watched people not know what to open. I emailed them and got it won't open back. I screenshotted them, which throws away the tabs and the slider and the whole point. A file you cannot share is just a fancy draft. So we built the step that was missing.

How to actually start

You do not need to learn anything first. You need to break one habit.

Stop asking for the answer as text. Ask for it as an artifact, and name the artifact. That single change is most of the difference between a weak prompt and a strong one.

Three things sharpen almost any of these. Name the format. Name the sections. Name one thing you want to do with it, a control to move or a color for the parts that matter. Build the prompt below and see.

That is the whole skill. Not syntax. A better request.

I did not learn to code this year, and I am not going to. I learned to ask for the right thing, and to put it somewhere people can use it. That was enough to make a hundred reports my team actually reads, and to help build the product that hosts them.

I just stopped asking for text.