I generate many html artifacts. I use the artifacts for many different use cases, like learning about different companies and technical concepts, and playing simple games. I like html because it’s easier to read than a wall of text.
Artifacts would typically be fragmented and hard to access, after generating from an agent, like Claude Code.
I built html-hub.com to reduce friction for myself. For uploading, Claude Code can figure out how to upload directly. The site supports directories and auto-selecting, or creating a new directory on upload. It also supports citations, which is derived from agent transcripts used for generating the html artifact.
This article will talk about why I built out my own public html library.
One topic I was curious about was Prometheus, Jeff Bezos’s latest company.
The company launched about 8 months ago without much disclosure to the public about the direction. He recently gave an interview, sharing more around Prometheus building a modern CAD.
I asked Claude to research what Prometheus did and to distill it into an HTML file. It generated a HTML report that made it clear how Prometheus’s product would accelerate the manufacturing loop.
Although I’m not a manufacturing professional, these explanations make it more clear how a Claude Code for CAD could accelerate the “idea to manufacture” loop.
An artifact like this would typically just get stuck in my file system, hard to track down and annoying to open. By building out html-hub.com, this enabled me to manually upload it via the UI and get back a public link.
I wanted the link to be public because 1) I wanted to easily share the findings with friends, and 2) I wanted my coding agent to have easy access to the information so it could build upon it for future research reports. Note that the content for the links are immutable too. This means that a person can reference the link without having to double-check its content because it’ll never change.
After the initial research of what Prometheus seemed to do, there were other questions that naturally followed. I asked about long timelines that Prometheus could potentially help with.
These are some long timelines. Also, one of my favorite things about agents producing HTML are the data visualizations.
Because Prometheus has raised $18B to date, one may grow curious about how they’ll price the product. In this report, it’s suggested that per-seat pricing could be used, though the pricing that would justify the valuation is “outcome & royalty” based pricing.
Crazy to me this is all information that we can gather all by asking some questions in a chat box. It’s not guaranteed to be correct of course. AI agents do seem to be a force multiplier on people’s skills though. For example, some investors have demonstrated they’re able to correctly infer a lot from public information (not saying this about myself).
One way to give more insight into how a html-hub.com report is created is by attaching a transcript. The main reason I wanted to add this is to know which of my HTML reports are cited the most, which would give an indication of value. In the transcript of the Prometheus pricing report, we can see two other html-hub.com links mentioned in the initial user prompt to Claude Code.

Agents can learn how to upload a transcript via https://html-hub.com/llms.txt. The LLM instructions give a suggested format for the transcript. Because the instructions ask for html-hub URLs explicitly when referenced, this allows html-hub.com to parse the transcripts, and determine which report is cited and how many times. It’s reminiscent of how the internet determined value for web pages when search engines were first getting started.
You might be asking, “Do you really drag and drop all these html files?” The answer is no. I ask my agent to do everything: research, create html, upload html, add html to digital library. The prompt orchestrates everything.
In the following image, we see the agent finishes all the work asked from the prompt.
It’s phenomenal how much cloud agents can do with the right tools.
By using cloud agents, I can kick off a research job, close my laptop, and return to read it when convenient.
I wrote this to share some of the fun insights I got from looking into Prometheus and building my own public HTML library. Friends enjoyed learning about the insights too. This project is experimental. If you have any thoughts, would love to hear them.
I will now share some other features I enjoyed building out for html-hub.com.
Autosort
When I add a html-hub.com link with a title, the report will either get added to an existing directory or a new directory will be created.
Having the links organized by directory is easier to navigate for readings than navigating through all my Claude Code web sessions.
This feature could use some improvement because right now it’ll only sort the added item based on title and not html content. This means a html item may get added to the wrong directory some times, and there’s no resource through the UI right now.
Latest Activity
Because it was difficult to know what the latest html pages getting added were, I added a “latest activity” section.
AI Chat
Asking questions with the html content.
HTML preview
On hover, you can preview the HTML before clicking on it.
Games
Not as much a feature, but rather a fun exploration. By creating HTML games that are public and immutable, it’s easy to create variations of existing ones.

And that’s a wrap for now. Thank you for your time.











