The AUA AI Hub Has a New Design¶
September 28, 2026
The AUA AI Hub has been redesigned. Most pages say what they said before; what has changed is how they are arranged and how you reach them, and those changes were made against measurements rather than by eye alone.
What you will notice:
- Navigation in one place. The menu columns down the sides of the page are gone. Every page has one control in the bottom-left corner: Browse opens a map of the section you are in, and On This Page jumps to any section of the page. After a few seconds it folds into a small compass labeled Navigate; hover over it or tap it to open it again.
- The full width of the screen. Reading pages now use the whole screen on a desktop, with sections side by side in outlined panels instead of one narrow column, and lines kept to about 80 characters.
- Colors that mean something. Each color stands for one kind of page everywhere on the site (tools, prompts, guides, lessons, benchmarks, news and events, governance), and a small color key appears wherever colors are used.
- Pictures of Antigua. The main pages carry scenes of the island drawn in code from photographs, under a night sky based on a real May evening over Antigua, with some artistic license. They are decoration only; beside the page titles they are left out on phones, and the home page keeps its view of the coast.
Why¶
The site had grown to more than a hundred pages, and its navigation had grown with it: one menu tab held 26 of the site's 64 entries. On a desktop screen, reading pages averaged about half their space in large empty areas, and on a phone some pages ran to more than sixteen screens of scrolling.
How It Was Done¶
The changes were tested against how the site is used:
- Everyday tasks. Artificial intelligence (AI) agents stood in for visitors (a first-semester student on a phone, a student on rotation, a course director, an administrator and others) and worked through tasks on the old site and on each proposed navigation, with the clicks and the backtracking counted. Some tasks were held back from the designers, to test whether each design also worked on tasks it had not been built around.
- Designs compared on one measure. Several page layouts were built in parallel and measured by the same script, above all for how much of each screen is left in large empty areas and how long each line is.
- An independent reviewer. A second AI model, from a different company, reviewed the designs, and its critique changed the outcome more than once. The final choices were made by the site's maintainer.
Some of the results, measured at different stages of the redesign:
| Measure | Before | After |
|---|---|---|
| Reading pages on a desktop screen, average share in large empty areas | 51% | 19% |
| Prompt Library for students, screens of scrolling on a phone | 16.1 | 5.6 |
| Home page, screens of scrolling on a phone | 6.7 | 4.5 |
The test visitors were AI agents, not people, so the real test is your use of the site. The full account of how the redesign was done, including what did not work and what was overruled, is in the Project Write-Up When Every Design Passes.
If something is harder to find than it used to be, please say so: the feedback form takes about two minutes, and contact details are on the About page.