The internet of intelligence, the launch and where to go next

The internet of intelligence is a network where AI agents act beside people and a site becomes a conversation and a service; a direction, not a fact.

What you can do now

Look back. You began with how the internet works and came to a network in which AI agents walk beside people. On the way you built your site eight times: with tables, with a separate stylesheet, in WordPress, responsively, from Markdown, on a headless CMS, with an AI builder and finally through an interview. You know how DNS and a certificate work, what a request and a response are, how a browser draws a page, why data and rights matter, how to write for a screen, how a story and persuasion are built, how to measure and how to check someone else’s output.

This set goes by different names, but its essence is one: you understand what a site is made of. And that, as this whole part has shown, is what separates a person who directs an AI from a person who believes in magic. The magic ends at the first invented number, the first key in the code, the first form with no privacy notice. Understanding doesn’t.

“The last textbook of the pre-AI era”

The authors of this textbook call it the last textbook of the pre-AI era. This is a working formula, not a prophecy. It means this: the textbook was written when AI had not yet become an ordinary tool for any work on a site, and so it teaches everything, from the very start. From now on people will learn and build together with AI: models will write what you used to write by hand. But a person still needs to know what to ask for, how to check it and what to answer for. That is what the whole course is for.

It doesn’t mean that professions will vanish, that AI will do it all itself, or that there will be no more textbooks. Nobody knows exactly how work on sites will change, and all that can honestly be said is that it already has.

The internet of intelligence: a direction, not a forecast

A working name for where this may be heading is the internet of intelligence. The idea, in a few lines.

• From pages to conversation and action. People more and more ask and delegate instead of searching and clicking. An answer is put together from many sources, and an agent takes the action on a person’s behalf.
• From a site to a service. A site as a set of pages is joined by what it can do: book, choose, calculate. This is what you decided in the lesson on agents: which doors to open.
• People and agents in one network. The agents of projects, people and services make agreements with each other (A2A, MCP). Trust grows in importance: who is this, what may it do, does a person confirm.
• A lot generated, little checked. When a page can be had in a minute, the value moves to what can’t be generated: living people, facts that can be checked, real work, responsibility.

All these are directions, not guarantees. Which protocols survive, which products vanish, how far agents become visitors: none of that can be forecast honestly. A year ago this textbook didn’t have half the names from the last lessons. So instead of a forecast, a more useful question is: what stays valuable in any scenario?

• Clean meaningful markup and accessibility. People, search engines and agents all need it. No scenario makes a page of div better than a page with real headings.
• Your own domain and your own data. What you own outlives changes of builders, platforms and fashionable words.
• Honest facts that can be checked. The more noise, the more precious what can be verified.
• The ability to set a task and check the result. This is the main skill beside an AI.
• Understanding. It carries over to any next tool.

Check yourself: why is the “internet of intelligence” called a direction and not a forecast?

Because it is unknown which protocols and products will survive, and how far agents will become visitors. You can talk about lines of movement and about what is lasting, but not about timing and forms.

The prologue, but now yours

In the prologue you saw your site pass through the eras, one after another. Then it was a show. Now it is your work: eight versions lie in the gallery side by side. Move the slider: the words and the main action are the same in every era, and only the clothes and the tools changed. That is the answer to the question of what in your work stays the same.

Era: 1996

My project

The same words in every era. Only the clothes changed.

Get in touch

1996: Tables and the font tag

• Tools: HTML tables, a GIF counter, FTP.
• What changed: For the first time a page could be yours: anyone could publish their own.
• What stayed: The text, the link and the name: what people came for.

The prologue promised that your own site would pass through the eras. You made eight versions of your project, and in the gallery they lie side by side. Look at them not as a museum but as an answer to the question: what in your work has stayed the same?

The launch checklist

Before you show the site to people, go through the list. It is gathered from the whole course: HTTPS and the domain, accessibility, meaningful markup, facts, the form and the law, outside scripts and keys, speed, link previews, backups and monitoring, the counter and cookies. For every item there are three honest answers: “yes”, “not needed” (if the item isn’t in your project) or a date for the fix (“by Nov 15”). An item that is “no” with no date is a red item: the site isn’t launched with it. It is better to launch later than with a lie in the list.

The first thirty days, and the main metric

After the launch the site isn’t finished, it is only beginning. Choose one main metric, tied to the project’s main action: for example, the number of inquiries for the action “Book a consultation” a week. The number of visitors with no inquiries is a vanity metric: it grows while the business doesn’t. Set a target number and look at it every week.

The plan for the month is simple. Week 1: publish, set up the counter, check the main action on a phone. Week 2: go through the page with the Tab key, read the text aloud, fix what you find. Week 3: look at where people come from and where they leave. Week 4: write a first piece on the questions visitors asked you. Don’t overcomplicate: a simple plan you carry out is better.

Where to go next

The course covered a lot and exhausted nothing. Choose three topics where you wanted more depth or that pay off for your project. Some signposts.

• Metrics and behavior, if you want to understand what happens on the site.
• SEO and getting into AI answers, if you need visitors from search.
• Law and personal data, if you have forms and clients.
• JavaScript and your first serious project, if you’ve felt like writing code yourself.
• A design system, if there will be several sites.
• Text, copywriting and stories, if you want the site to persuade.

Where to go for material. MDN’s Learn web development and web.dev/learn are open courses from the makers of the standards and the browsers. The Modern JavaScript Tutorial (javascript.info) teaches the language itself in depth. Harvard’s CS50x is an open course on the basics of programming. And read the documentation of any tool at its makers’: it is more exact than any retelling, this one included. All the links are in the sources, and you can come back to them from any lesson.

And one last thing. The textbook is open: the texts are under the CC BY-SA 4.0 license, the code under Apache-2.0, and the repository is on GitHub at NoeticEcho/learn. It is a NoeticEcho project; follow it at open.noeticecho.space. If you find a mistake, an inaccuracy or an out-of-date date (and in fast-changing fields there will be some), report it: it is the only way to make the book better.

Practice

Draw the conclusion for your project. Go through the checklist of twelve items and enter an honest status for each: “yes”, “not needed” or a date. Name the main metric with a number and lay out four weeks. Choose three topics to study.

The check reads your record: it will make sure every item has an allowed status (a red one with no date won’t pass), that the metric is tied to the main action, that the plan is four weeks and that there are three topics. Nobody can check the honesty of the answers except you. If the site isn’t ready yet, write down the dates and come back. The course will wait.

What was wrong with 2025

In 2025 it seemed that learning how a site is built was no longer needed: an AI would do it all. By the end of the year the opposite had become clearer: those who understood what to ask for and how to check got sites they could trust, and the rest got fast pages with invented numbers. Speed became cheap, and understanding stayed expensive. This book taught the second.