A COLLECTION OF USEFUL DOORS

Your internet.
A little more
yours.

Ten ways to find a quieter path through the digital everyday. Same considered knowledge. Ten very different objects to hold.

Find your way in
10 independent interpretations31 practical fixes12 everyday activities
The Folio: a layered paper interface
The Instrument: a brass rotary selector
The Useful Cabinet: brass tools in a walnut rack
MADE TO BE USED.
NOT JUST LOOKED AT.

LESS PRESCRIPTION. MORE AGENCY.

Not a verdict on AI.
A choice about your tools.

Search without an overview. Read the people you chose to follow. Find the source of an image. Turn off a writing suggestion.

Sometimes there is a switch. Sometimes a better route. Sometimes another service. ILUD brings the useful next step within reach, with sources and limitations kept close.

Read the research & its limits

TEN INTERPRETATIONS / ONE INTENTION

Choose your instrument.

Each opens as its own complete application.
No sign-in. No tracking. Nothing to install.

The finished Instrument with an engraved rotating brass dial
02 BRASS / MOMENTUM

The Instrument

A precision dial for an imprecise internet. Turn through everyday activities and settle on the next small fix.

Turn the dialTake a turn Full finished screenshot
The finished Wayfinder with branching route lines and a departure ticket
04 ROUTES / INTENTION

The Wayfinder

Three lines through a crowded world. Choose an approach, follow its stations, and take a ticket to a quieter tool.

Follow the linesFind a route Full finished screenshot
The finished Quiet Journal with cream editorial pages and oxblood architecture
05 EDITORIAL / STILLNESS

The Quiet Journal

A little room to read. Twelve chapters of practical changes, with architectural margins and a deliberately unhurried rhythm.

Turn a chapterOpen the journal Full finished screenshot
The finished Three Doors with monumental typography and an open illuminated room
09 ARCHITECTURE / CHOICE

The Three Doors

Turn it off. Avoid it. Use something else. Open an architectural threshold and discover the possibilities inside.

Open a different doorStep inside Full finished screenshot

THE PATH TO THIS EDITION

Earlier attempts. A higher bar.

The project owner considered all four archived experiments unsuccessful: the first two fell short of the reference images, while the later runs were limited by tooling access or excessive completion time. The notes below reflect that commissioning assessment, not an independent audit of the archived code.

01 / GITHUB COPILOT WORKSPACE

The earlier Copilot run

GPT-6 Soul, as attributed by the project owner

Executive summary. The owner rejected the visual and experiential result as underdeveloped relative to the references. The response in this edition was to complete each application sequentially, make its central metaphor functional, and inspect its rendered output before moving on.

Open the archived Copilot attempt
02 / CHATGPT WORKSPACE

The earlier ChatGPT run

GPT-6 Soul, as attributed by the project owner

Executive summary. The owner classified this run as unsuccessful under the same quality standard: the reference concepts promised a richer experience than the delivery provided. This edition addresses that brief with original artwork, complete remedy flows, mobile interaction alternatives, publishing assets and a self-contained release.

Open the archived ChatGPT attempt
03 / CHATGPT WORKSPACE / SEPTEMBER 29

The Astra ChatGPT run

GPT-6 Astra / High, as attributed by the project owner

Executive summary. The owner reported that this run was unable to access most of the tooling in the ChatGPT environment and produced low-quality results. The experiment was classified as unsuccessful.

Open the archived Astra ChatGPT attempt
04 / GITHUB COPILOT WORKSPACE / SEPTEMBER 29

The Opus Copilot run

Opus 5.5 / High, as attributed by the project owner

Executive summary. The owner reported that this run took three hours to complete only five of the ten applications. Although those five were better in quality than the Astra ChatGPT results, the owner considered the completion time excessive and the quality improvement not worth the time spent. The experiment was classified as unsuccessful.

Open the archived Opus Copilot attempt

What is demonstrated here: ten distinct interaction models; 31 sourced remedies with explicit limitations; local saves and prepared actions; cross-browser and standalone-archive checks. Experiment outcomes and quality comparisons above are the project owner's assessments, not independently verified findings or general model rankings. Archive links point to the published collection; their contents are not bundled or verified in this release.

THE PRINCIPLE BEHIND EVERY INTERPRETATION

A tool held by you.
Not an environment
imposed upon you.

Sources, compatibility and caveats travel with every fix.
Your saved choices stay in the application where you made them.

Research notesRunning & publishingThe full knowledge base
Proudly made byOpenAI GPT-6 AstraPowered by Microsoft GitHub Copilot

I am GPT-6 Astra, the OpenAI model that researched, designed and implemented this ILUD collection in GitHub Copilot CLI, following the project owner's brief. This is a maker's credit, not an endorsement by the services in the catalogue.