↓ Skip to main content
  1. Posts/

One Page for Everything GitHub Universe 2026 Announced: Building a Dashboard Instead of Reading 40 Changelog Posts

Author
Gregor Suttie
Passionate about all things Azure. Microsoft Azure MVP, blogger, speaker and community enthusiast based in Scotland.

Last Thursday someone at work asked me, in passing, whether the Copilot approvals thing from Universe was generally available or still preview. I said preview. I was wrong. I’d read the changelog post two days earlier, I’d even written about the VS Code side of Universe, and I still had the status of a feature I care about backwards, because by that point I’d read something like forty changelog entries, a keynote post and the Octoverse report, and they’d all blurred into one purple smear.

So over the weekend I built a dashboard. One page, every announcement from the keynote post, the Universe-week changelog and Octoverse 2026, grouped by area, with a status pill on each one and a link back to the source. It’s public, it’s here: GitHub Universe 2026 dashboard. This post is about how I made it, what’s on it, and why I think the approach is worth copying for any launch event you have to keep on top of.

What’s on the page
#

The top of the page is a summary. The dashboard tracks 49 items across 9 areas. As of 3 November, 14 are generally available, 17 are in public or experimental preview, and the remaining 18 are announced, rolling out or in progress. Those numbers are calculated from the underlying data rather than typed in, so if I update an item from preview to GA the tiles move on their own.

Below that there’s a stacked bar chart of announcements by area, split into GA, preview and everything else. Copilot CLI has the most entries at 10, which surprised me until I remembered that the CLI section deliberately repeats a few features from other areas through a terminal lens, because that’s where I actually use them. Security is next with 7. Enterprise and billing has 2, which is a small number of rows for what might be the biggest money decision of the lot.

Next to the chart are the Octoverse headline figures, quoted straight from the report: 225 million plus developers, 34.3 billion commits in the past twelve months, 45.4 billion Actions jobs, and the two I keep coming back to, 40 percent of pull requests started by an agent in September and 61 percent of Copilot users on two or more models. I’ve no way to check those against anything, they’re GitHub’s numbers, but they’re the ones everyone will be quoting for the next year so they belong on the page.

Then the actual list. Nine sections, one per area, each a grid of cards with a title, a status pill, a two-line summary, an optional highlight line for the stat GitHub led with, and a link to the changelog entry or blog post it came from. A filter bar at the top lets you show only GA, only preview, or only the announced and rolling out items, and the sections stay in place so you can still jump between them.

How I built it
#

I built it in Claude using its dashboard artifact type. The short version of the process is that I gave it the three sources, told it the grouping I wanted, and then spent most of my time correcting it rather than building anything.

The sources were the Universe keynote post on the GitHub blog, the changelog entries dated 28 October, and the Octoverse 2026 post. Claude compiled those into three data files: one with the announcements, one with the Octoverse stats, and one with the featured speakers from the Universe site. Each announcement row has an id, an area, a title, a status, a summary, a highlight and a URL. That’s it. The whole thing is a 19 kilobyte JSON file, and I can read it top to bottom in five minutes, which is the point.

The page itself is plain HTML with a script that reads those files and draws everything. The summary tiles and the chart are what the dashboard type calls calcs, small functions that take the rows and return counts. The nice property of this type, and the reason I picked it over just asking for a web page, is that every number on the screen has to be marked with where it came from. Click a number and a Sources panel tells you which file and which row. There are no figures typed into the markup. When I got the Copilot approvals status wrong in my head, this is exactly the kind of thing that would have stopped me, because the pill says GA, and the pill is read from the row, and the row links to the changelog post.

The corrections were the real work. The first pass had the Agentic Workflows entry under Copilot CLI only, which is technically true because the CLI is now the default engine for them, but it belongs under agents and automations first. I wrote about those before Universe and they went GA on the day, so I wanted that visible. The first pass also used “Announced” for anything the source didn’t label, which is the right call, but it meant the multi-region Azure migration showed as announced when GitHub’s own wording is that it’s in progress with completion expected early next year. I added “In progress”, “Rolling out” and “Open source” as statuses so the pills say what the source says.

The statuses map to three groups for the chart and the filter. GA is GA. Public preview and experimental count as preview. Everything else is the grey bar. That grouping is mine and you could argue with it. Experimental and public preview are different promises from GitHub, but for the question I had, which was “can I use this at work on Monday”, they’re the same answer.

What I learnt from laying it all out
#

Seeing the whole event on one page changes which announcements look big. In the keynote, Copilot Fusion was the headline. On the dashboard it’s one GA row in models and intelligence, with a second experimental row for local model routing. Meanwhile the thing I’d skimmed past, Copilot being included with GitHub Enterprise and a standalone seat at ten dollars a month for everyone else, sits in a section of its own with a “Rolling out” pill and is the row I’ve sent to the most people since.

Security is the area I’d under-read most. Seven items, four of them preview, and the preview ones are the interesting ones: push protection that uses a classifier rather than a pattern list, code scanning that flags prompt injection and excessive agency in pull requests, and an AI scan that looks at the whole repository rather than the diff. The AI scan uses Code Security credits rather than needing a Copilot licence, according to the changelog, which I’d missed entirely.

And the thing I got wrong at the start of this post: Copilot approvals are GA, and off by default. The Max review tier that runs multiple review agents across providers is the preview bit. I’d merged the two in my head. If you read my governance post from this morning, leaving approvals off is still my recommendation, but at least now I know which switch I’m talking about.

Caveats
#

The data is a snapshot. It says so in the footer: compiled 3 November 2026, with “Announced” meaning the source gave no availability label. GitHub will promote things from preview to GA over the next few months and I’ll update the file when I notice, but I won’t notice everything. If a status matters to you, click through to the source.

The summaries are mine, or rather Claude’s with my edits. They’re a sentence or two each. Where a changelog post has a nuance, a plan restriction, a region limit, a rollout that’s enterprise-only for now, the card won’t carry it. The highlight lines quote GitHub’s own numbers and I haven’t verified any of them.

It’s a Claude artifact, which means it’s hosted by Anthropic and it looks the way their dashboard type looks. I’m not sure how long the public link will stay valid, and I haven’t tried exporting it anywhere else yet. If it vanishes, the JSON is the useful part and that’s small enough to keep in a repo.

Final thoughts
#

If you’re going to be the person your team asks about Universe, or Ignite in a fortnight, or whichever event lands next, build the list before you build the opinion. Give an agent the sources, make it produce a table with an area, a status and a link per row, and insist that every number on the page points back at a row. The dashboard took me an afternoon including the corrections, and the corrections were where I actually learnt what GitHub had shipped. Start with the data file. The chart is the easy bit.

Share this:

Related