New site, who dis?

As I do every year, I went through my personal site to do some refreshing and improvements. Here are the main takeaways and lessons learned, in case you are curious.

So, let's start with the design changes.


In terms of design no much has changed, I stuck with a minimalistic style. As a small site, scaling isn't a threat and it helps to group tokens and decisions with the mind set on performance and minimal style foot print.

I moved the site to a new repository, which means you can still check out the old one and compare it if you want to.


As said before, though styles haven't changed a lot I coded all from scratch and there are good reasons for this.

In my previous version of the site I was over-adjusting browsers default styles using bare tags as style selectors and using some elements not for their semantic meaning but because of the styling shortcut they provided.

Tags for content, class names for styles

The methodology I went for is to only use class names and override browser defaults when really necessary. By doing this I focused on choosing the right and preferred tags for content, improving accessibility and CSS specificity. I'm also using CSS variables now for value distribution instead of LESS variables.

I'm still using LESS but as a bundling tool. I generate one specific entry file for each page, and each page has a type variable in the front matter to know which styles to pick.

I pass the result through cssnano to optimize it, export it as partial and later inline the styles in the head of the page.


Using CSS variables makes easier to share values without thinking of global LESS imports and enables a straight-forward dark mode implementation.

Inlining the styles reduces extra network calls and render blocking resources, but they also enlarge the size of the final HTML and doesn't take advantage of caching.

This is why having one specific stylesheet per page with only the rules needed and an aggressive minification are necessary and makes the HTML size footprint plus the no-cache strategy convenient.


There's no much JavaScript around the site, and most of it takes care of the web font loading strategy.

The static output assumes the user has disabled JavaScript and unless an inline script runs at the beginning then no animations take place on the homepage, and no web font or other dependencies are loaded. The site is progressive and as you have a better network and better browser is applies further optimizations and features.

Font loading strategy

If the user has JavaScript enabled then, a stylesheet containing font face rules and a script which will observe the fonts are both asynchronously loaded.

After fonts are ready store-css saves font face rules rules in web storage along with a fonts-cached flag for future inner navigations.

I wrote about this approach in this article a while ago if you are interested.

Dark mode and prefetching

The rest of the JavaScript in the site handles the dark mode toggle logic also using web storage and to prefetch links the user might visit.

The later is achieved by combining IntersectionObserver and link[rel=prefetch] both present in most modern browsers.

When any of those aren't available or the user is on a slow connection then the scripts do nothing about it to avoid loading unnecessary polyfills or consuming user's data.


The home page has some orchestrated animations. The page waits for a maximum of three seconds for web fonts to be loaded, if they don't animations will kick off either way or won't run at all when JavaScript is disabled.

In the previous version of this site I was using transitions and transition delays, this caused them to be fast-forwarded on repeated views and strange side effects on resizes. This is why I moved to animation keyframes.

Static generator

Originally this site was using Jekyll and GitHub Pages, a really convenient setup to quickly launch something without much overhead and a solution I would still recommend to anyone.

Sadly as the content grew, Jekyll builds became super slow, making my writing experience a bit frustrating, so I decided to give Eleventy a try.

Eleventy is basically a Jekyll implementation in Node.js, but easily extensible and more template language options. The most important thing is that builds are really fast folks.

New ecosystem, new gotchas

The move was smooth in a lot of corners, and not as smooth as I would have liked in others. Most of the package defaults are the same as in Jekyll, like template language and directories which it's great when migrating.

But now liquid and markdown parsers are JavaScript implementations which have differences with their Ruby counterparts so some snippets weren't just working out the box.

Most of these differences are not bugs though, since there aren't real standards around but is good to keep these gotchas in mind if you are going to move a Jekyll site to Eleventy.

In addition, Jekyll extends liquid with some useful filters that aren't present, the good thing is that Eleventy is super configurable around this so you can create these missing parts in seconds.

Eleventy has really good up-to-date documentation and Paul Lloyd wrote a great introduction article I strongly recommend.


One of the things performance tools were penalizing my site for was low server response which wasn't much in my control when I was using GitHub Pages. Now, this site lives in Netlify supporting secure connections, branch previews and continuous deployment in addition to CDN distribution.

Netlify is the thing I wished was available when I was a kid throwing invalid HTML to an FTP server like twenty years ago

I always recommend to go for the simplest setup to new developers and beginners to focus more on code and learning than in infrastructure and servers, and Netlify is just so simple I took my own advice, and it was the easiest thing to tackle from a the long migration list I had.


As I said on Twitter some time ago, a personal site is this place where you can show who you are and by roaming around mine and reading this piece you will notice that, first, I'm definitely not a designer and second, I'm quite obsessed with performance and content first delivery.

In my carrer I've seen how business was the main threat to a building a web product with good performance, third-party scripts, contradictory product decisions and badly orchestrated feature implementations makes it hard to think of a global and future proof strategy around best practices.

This is why I found interesting to hear stories on how teams dealt with optimizations while responding to business priorities but here there's no business, this is me and my playground. My playground, my rules.


Huge thanks to Zach Leatherman for Eleventy, Jun Yang for liquidjs, and Alex Kocharin and Vitaly Puzrin for markdown-it, which this site heavily relies on.