Building a Dashboard for NickOS

Overview

I recently built a dashboard for NickOS, my local AI operating system and the repository for my various knowledge databases.

It started out as a simple weekend project to build an aesthetically appealing landing page, centred around a constellation-like particle field, for the links I use often but never seem to have in one place, especially when moving between different devices and browsers: company intranet pages, administrative forms, shared working documents, internal guidance and Notion databases.

As the project grew, it also became an exercise in using the right model for the right task. Because Fable 5 is capped at 50% of weekly usage through 7 July, I began using it more sparingly and deliberately for planning, with Opus 4.8 and Codex handling execution.

What It Grew Into

After several rounds of iteration, the dashboard evolved into more than a link page. Its panels read directly from the local files NickOS already keeps, so the page simply renders what the system has already written or produced: the latest items from the morning automation that scans for gaming-industry regulatory developments, the newest candidates for my weekly work report, and the notes most recently ingested into my regulatory wiki.

The wiki also has its own panel: a searchable, rearrangeable index of everything ingested, sitting alongside generated insights. At a glance, it shows the kinds of matters I have been reading about or working on recently, while surfacing synthesised knowledge and ideas drawn across them. The dashboard also includes a customised Markdown viewer, styled with the theme, typography and formatting I prefer, so I no longer have to open a separate application just to read the .md files NickOS produces.

The result is a quick view of what has moved, what has been created, and what may require further attention. It also gives me a better reason to keep improving the wiki, because the material no longer disappears into a folder structure after ingestion.

Well, I am now more or less out of usage on both Claude and Codex for the moment, but it was a satisfying project to finish!

Back to AI Journal