HTML-ization
For my whole career, the distance between what I could see and what I could ship was the actual job.
“It has always given me satisfaction to be able to create what I have in mind.”
That sentence, or quote if you may, is the whole thesis of this edition. It’s stick to me for a while now.
Not the productivity. Not the speed. Not the stack.
The satisfaction of being able to build the thing you can see in your head.
For most of my career, that gap, between what I could see and what I could ship, was the actual job. Hiring filled it. Capital filled it. Time filled it. Sometimes none of those filled it, and the idea just died.
In 2026, that gap has collapsed for anyone willing to learn how to close it.
That’s what this edition is about. That’s what the thing I’m publishing this morning is about. That’s why I made the source open.
Five times a week, someone asks me the same question.
A DM. A text. The back of an Uber. The barista who recognizes the logo on my laptop.
“Where do I start with AI?”
Cringe, I know. But we have all been there. Or the slightly better version:
“How Claude Web, Cowork, and Claude Code are different”
For two years I gave the wrong answer. A YouTube link. A podcast. A “just try Claude Code and see what happens.” Each one was a fragment of a fragment.
The honest answer was that I would have to write a book. Or a playbook. Or a course. I still don’t know what it is, and that turns out to be the point, I’ll come back to it. Everyone who’s read it tells me it’s the most useful thing I’ve put out, so let me just frame it like this:
So I wrote IT.
This morning, the repo went public. More than forty chapters. Free. No signup. The source is on GitHub.
The whole thing took thirteen days and a swarm of more than twenty-five agents running in parallel, built on my knowledge base, my data, years of docs, and the workshops I ran in the past.
Before you click anything, though, you need the one idea that made it possible in the first place.
Because this isn’t really an essay about a playbook.
It’s about how operators ship in 2026, and why that finally feels like the thing in my head.
The gap
Every idea you have ever had lives in two places.
There is the version in your head. Sharp, complete, exactly right. You can see the pitch, the product, the page, the company. You can almost touch it.
And there is the version you can actually ship by Friday. Smaller. Compromised. Shaped less by your imagination and more by what your tools, your budget, and your own two hands will allow.
The distance between those two versions has a name I never used until this year.
The Imagination Gap.
It is the single most expensive thing in any operator’s life, and almost nobody prices it.
For most of my career, closing that gap was the job. I did not get paid to have ideas. Ideas are cheap; I have a notebook full of them, and so do you. I got paid to drag an idea across the gap. To translate the thing in my head into something real, through the only instruments available.
Hiring was the gap-closer. You can’t build it, so you find people who can, and you spend a year and a salary doing it.
Capital was the gap-closer. You can’t afford the version in your head, so you raise money to rent the version you can afford.
Time was the gap-closer. You can’t do it now, so you queue it, and you watch it sit in a backlog while the market moves.
And when none of those were available, the gap simply won. The idea stayed in your head. It died at the dinner table. You told it as a story years later: “I had that idea before they did.” Of course you did. Everyone has the idea. Having the idea was never the rare part.
Crossing the gap was the rare part. The gap was the filter.
What the gap was actually filtering
Here is the part almost nobody says out loud.
The Imagination Gap did not just slow ideas down. It selected them.
A wide gap is expensive to cross, so you only cross it for ideas that already look safe. The big, strange, ambitious idea, the one that needs a thing that doesn’t exist yet, that one stays home. Too expensive to test. The cheap, obvious, incremental idea gets built, because it’s the only one the gap will let through without a fight.
So the gap was never neutral. For decades, it quietly decided which ideas got to exist at all. Not the best ideas. The cheapest-to-cross ones.
That is why the speed conversation about AI bores me.
Everyone wants to talk about AI making you faster. Faster is the small read. Faster just means you cross the same old gap a bit quicker, building the same cautious things.
The real shift is not speed. The real shift is that the filter is gone. When the gap costs almost nothing to cross, the ambitious idea and the safe idea cost the same to test. Which means the strange one finally gets built. Not because you got faster. Because the thing that used to kill it is no longer in the room.
AI did not make me a faster operator. It changed which of my ideas are allowed to be real.
AFC?!
Friends. Steak. A dumb idea I couldn’t shake.
An Autonomous Fighting Championship. Humanoid robots in an octagon. Blacklight, bass, no humans bleed. I said it out loud and watched the table decide whether I was serious.
The question that came back wasn’t “is this insane?”
It was “Do you have a deck?”
The honest answer was no. There was a thought and a stake.
Five years ago, that is exactly where the idea died. The gap was too wide. To get a deck good enough to send, I’d need a designer, a few hundred dollars, and two weeks I didn’t have for an idea I wasn’t even sure about. So it stays a dinner story. “I had this idea once.”
Instead, the deck existed before the next meeting did.
Not a slide file. An interactive pitch. A use-of-funds donut you can hover. A revenue model that toggles by year. A companion piece with every humanoid robot platform shipping in 2026, filterable, with a profit-per-fighter calculator where you move the sliders and watch the math move.
It looks like the thing that was in my head. Not a compromised, gap-taxed version of it. The thing itself.
I sent it around. It circulated. AFC might become a real company (jokes aside). It has a non-zero chance for one reason only: the idea was allowed to leave my head intact, the same week I had it.
That dinner didn’t teach me anything about robots.
It taught me the gap was gone, and I hadn’t noticed yet.
What closing the gap does to the work
When the gap is wide, you spend your life translating downward.
The vision is a 10 → Your tools allow a 6 → So you ship a 6 and quietly grieve → the 4 you lost in translation.
Every operator knows that grief. It’s the slide that didn’t land the way you saw it. The product shipped 70% right.
When the gap closes, the translation stops. You stop asking “what can I realistically make?” and start asking “what do I actually want?” Those are completely different questions, and the second one is the one your imagination was built for.
Three things changed in my own work the moment I stopped translating downward.
The artifact stopped being a compromise.
A PDF is what’s left of an idea after reality taxes it. It’s static because static was cheap. An interactive artifact, a thing you can click, sort, filter, and play with, is what the idea looked like before the tax.
The medium stopped subtracting from the vision.
The categories stopped fitting.
Remember, I said I don’t know if I made a book, a playbook, or a course? That’s not me being coy. Those three words are old.
They describe artifacts shaped by old constraints. A book is what an idea becomes when printing is expensive, a course is what it becomes when teaching has to scale through video.
The thing I made is a living site with forty-plus chapters, twenty-five interactive widgets, a tier list that updates itself, and three real case studies embedded clickable inside it. There isn’t a clean word for that yet. There isn’t a word because the constraint that would have shaped it is gone.
The iteration loop collapsed.
Six full editions in thirteen days. Not because I’m fast. Because when the artifact is alive, you fix it with a change, not with a re-export, a re-attach, and a re-explain. The format stopped fighting me.
HTML-ization
This is the one I want you to actually steal.
Once I noticed the gap was gone, I stopped reaching for the document by default, and I also converted my team to the same. The entire company is now doing HTML artifacts to make it more fun, new, and interactive.
The board update did not want to be a slide. It wanted to be a page you could click into, with a header, pages, etc.
The client audit did not want to be forty pages of charts. It wanted to be a thing where you sort the table, click a flagged row, and see exactly why it failed. The pitch did not want to be a static donut. It wanted a slider.
I started calling it HTML-ization. It is a small word for a real habit: when an idea is more than text, stop flattening it into a document and let it be what it actually is, an interactive thing.
Most of what we ship is interactive in our heads and dead on arrival. You picture the reader exploring it. They receive a frozen screenshot of that picture. HTML-ization is just refusing to do that last, lossy step. It is the difference between describing a machine and handing someone the machine.
And it is no longer a developer skill. That is the part people miss. Asking an AI to turn your messy thinking into a clean interactive page is now roughly as hard as asking it to write the email about the page. The gap that used to sit between “non-technical person” and “shipped an interactive artifact” closed at the same time as all the other gaps. Most people have not tried it yet.
The form used to shape the idea. Now the idea shapes the form.
And one honest footnote, because I know some of you are technical, and the rest are not. I happen to deploy mine to a link; I run a few companies, and a link is cleaner for me. You do not have to.
An interactive artifact forwards perfectly well as a file you attach, paste, or hand-deliver to someone. The link is my preference. It was never the point.
The point is that the artifact, however it travels, finally matches the thing in your head.
The receipt
I tested all of this against the biggest idea I had been avoiding.
For two years, the honest answer to “where do I start with AI” was “I’d have to write something about it.” That sentence was itself a confession of the gap. I’d have to mean the gap is too wide, so I won’t.
It’s done.
The thing is public this morning.
More than forty chapters on running AI as an operating system, the tool stack, parallel sub-agents, hooks, MCP, evals, voice agents, headless agents, browser agents, the failures that quietly explode your bill, and when to outgrow each tool. Twenty-five interactive widgets. Fifty-plus glossary terms linked inline. Nine dated research notes.
Absolutely free.
I just don't like all the people on LinkedIn, X, over email, and everywhere else selling this knowledge with “I did this with “that” prompt earned $$$” — it just doesn't make any fucking sense.
Free. No signup. No paywall. The only thing I ask is that you’re on this newsletter, so you see what comes next.
🎬 Post-Credit Scene
One pick per kind of reader. Things I’ve actually spent time with. Not a list copied off Goodreads.
📕 If you want the deep version of this whole edition.
The Beginning of Infinity by David Deutsch. His argument is that all problems are soluble, that the only real limit is knowing how. That is the Imagination Gap dressed in philosophy. Hard book. Worth every page.
🎙️ If you want to keep up without drowning.
Dwarkesh Podcast. The interviews with the people actually building agent infrastructure are doing more to shape how I think about this transition than any whitepaper this year. Start with a recent long-horizon agents episode.
📝 If you want the deeper “why” behind the playbook.
Andrej Karpathy on building an LLM-maintained personal wiki. The best thing I’ve read this year on what knowledge work becomes when the gap closes. Static knowledge bases are over. Living, queryable, self-curating ones are next.
🛠️ If you only install one tool this week.
Claude Code with sub-agents and the SKILL primitive. Read Chapter 13, then install it. The cost-to-output ratio is the most absurd arbitrage in tech right now, and most people are still using AI through a chat window like it’s 2023.
📺 If you want one show to watch while your agents work.
Pluribus on Apple TV+. Vince Gilligan’s new one. A story about what happens when the rules of a system change for everyone at once. Feels about right for the year we’re in.
Your call → dive.vladyslavpodoliako.com
Thanks for being here. Thanks for the replies, the shares, and the stars.
Vlad