Home /
I added a graph because the site needed some shape
A small site graph made billiem.uk feel more like a growing personal object than a fast but flat static site.
- The graph is a playful map of the site, not a breakthrough replacement for normal navigation.
- The live graph data shows 23 internal nodes, 174 external nodes, 302 edges, and the free-compute post as the largest link hub.
- Small LLM-assisted visual experiments can make personal sites feel less interchangeable when the cost of trying them stays low.
I added a graph view to this site because performance was not enough personality. The site was fast already. It had posts, tags, feeds, a sitemap, and the usual static-site plumbing. It still felt flatter than I wanted. Now it has a moving little map of itself, which is not the same as being more useful, but it is closer to the kind of web I want to make.
The idea was not especially deep. I remembered that sitemaps exist. I remembered seeing animated graph interfaces that made link structures feel alive. Then the thought was just there: the site should have one.
Two hours later, it did.
That is the part I keep coming back to. A lot of LLM-assisted projects still take ages because the first version is wrong in a way that needs careful untangling. This one did not feel like that. The first usable shape was close to the image in my head, and the remaining work was mostly tightening the behaviour until it felt good enough to leave deployed.
A map, not a better menu
I do not think this is a breakthrough in web navigation. I would not tell anyone to browse the site through the graph before using the homepage, tags, or RSS. Most of the time, a list is still the correct interface.
The graph earns its place differently. It makes the site visible as a structure instead of a sequence of pages. You can see the homepage, posts, tags, feeds, internal markdown alternates, and outbound references sitting in one moving canvas. You can search, fit the graph to the viewport, pause the motion, hide labels, toggle external links, filter by link context, and click nodes to inspect their URL, status, type, path, depth, and link counts.
That is more playful than practical, but it is not empty decoration. It shows what the site is becoming.
The live graph data generated on 30 June 2026 reports 23 crawled pages, 23 internal nodes, 174 external nodes, 302 edges, 407 weighted edges, and no crawl errors. Those numbers are small enough to be readable, which is probably why the feature works at this stage. The graph is not trying to visualise a mature archive. It is catching the site while the shape is still legible.
The star is the point
The funniest part is that the graph immediately made one post look ridiculous.
My free compute post dominates the view because it has a lot of external links. In the graph data, it has 174 outgoing links and 269 weighted outgoing links. Visually, it turns into a big star while the rest of the site is a looser web.
I both like and dislike that.
Part of me wanted to push the nodes apart more, tune the collision harder, and make the layout feel more even. But that would also be lying a little. The free-compute article is a big collection of links. Of course it looks like a hub. The graph is most interesting when it shows an awkward truth about the site rather than smoothing everything into the same decorative pattern.
That is where it becomes more than a toy for me. It gives me a way to watch the archive grow. If future posts become dense link hubs, sparse notes, tag bridges, or isolated little islands, the graph will show that without needing a dashboard full of metrics.
Personal sites should get weirder
I keep thinking about this as a small reaction against the flattened parts of the web. Social profiles, portfolio templates, and platform pages all tend to converge. They are orderly, legible, and interchangeable.
I am not pretending this site fully escapes that. The graph has a visible LLM-assisted aura. It is a canvas UI with glowing nodes, panels, search, filters, and a little too much polish in the places that current coding tools like to polish. It is not hand-rolled 2003 chaos.
Still, it points in the direction I like: a personal site as a place where odd features can exist because the owner wanted them there.
The graph does not need to justify itself as conversion design, growth design, or productivity design. It can just be a pretty map. It can make the site feel a bit more mine. That is enough.
This connects to what I wrote about LLM Choice: low-cost artifacts are useful because they create things to react to. The threshold for trying small ideas has dropped. Sometimes that means building a tool that saves money or replaces manual work. Sometimes it means adding a graph because the site needed some spice.
The rule I am taking from this is simple: if a small feature makes a personal site feel more alive and it can be built quickly without turning the rest of the system inside out, it is probably worth trying. It does not have to become the new interface. It can just make the place stranger in a way that feels true.