Bring your own
robot friend

For readers behind a border their favorite robot friend can't cross.

These collections are meant to be read with an AI — uploaded to a capable model that walks you through the work with adaptive footnotes and cinematic exploration. But not everyone can reach those models. Some regions are fenced off; some hosted models sit behind filters that flinch at a poem about a child in a youth hostel at the wrong place at the wrong time.

So here is another door. In about ten minutes, with a free account, you can stand up your own private reading on Cloudflare, running an open model. No invitation required. The reading lives in your browser and belongs to no one but you.

The collections
Collection III

Tao of Totality

a love letter to planet Earth
or: what happens when a Protector embraces the species

a space left blank — for now
Notify Me

I’ve wandered over to the display computers.

I giggle mischievously; first love notices.

·  ·  ·

“Let me guess,” he walks over.

“You’re installing Firefox on all the display laptops?

I’m surprised they don’t have that locked down.”

— an excerpt from Idle Hands

How to run your own reading

You'll download one file, create one free account, paste the file into a web editor, and click deploy. No coding required.

  1. Download the reader file for the collection above. It's one large text file — that's normal; you never have to read it.
  2. Make a free Cloudflare account at dash.cloudflare.com.
  3. Create a worker from a template. In the sidebar open Workers & Pages (sometimes labeled Compute) → Create application → Start with a template → choose LLM App. This switches on the AI model and wires it up for you. Name it, then Deploy.
  4. Paste in the reading. Open your new worker → Edit code. Select all of the starter file (it may be index.js or worker.js — either is fine), delete it, paste in the file you downloaded, and click Deploy.
  5. Open it and read. Click your worker's .workers.dev address. When the small dot near the top turns green, tap a door — “Adaptive footnotes, from the top” — and begin.
If something doesn't work · cost & privacy

The dot stays red. The model isn't connected. In your worker open Settings → Bindings and check there's a Workers AI binding named exactly AI. If not, delete the worker and redo the template step, choosing LLM App.

You see Cloudflare's old AI Chat page instead of the reading. This means the starter template's static website is being served instead of the worker script. The code preview can still look correct, because preview tests the script directly. The fix is to remove the template's static assets or site setting and deploy only the downloaded reader file as a single-file worker. Keep the Workers AI binding named exactly AI.

The worker URL won't load. In some regions *.workers.dev addresses aren't reachable. In the Cloudflare dashboard open your worker's Settings → Domains & Routes and add a custom domain you own; use that URL in place of the workers.dev one.

Cost. Light reading can fit inside Cloudflare's free daily AI allowance. Heavier use may need the Workers Paid plan (about $5/month). Start free; upgrade only if you hit a limit.

Privacy. Everything runs in your own account and browser session. Close the tab and the conversation is gone — no login, no history kept.

A gentle note. This runs on a different model than the one the collection was tuned with, so a reading here may phrase things differently or miss a reference now and then. The shape of the experience is the same.

I bite my lower lip

look up innocently.

·  ·  ·

“Open source represent,” I wink

moving over to the next laptop.