
The LifeLog story — how the blog that documents everything was born
The blog that wasn’t the center
In early July 2026, I had six active projects — Arachne, Dogwalk, Capivara, Portfolio, TatuEngine, and a growing list of technical discoveries. Each with its own repository, its own CI/CD, its own set of bugs and decisions. I knew tomorrow would have been better if I remembered what I learned today.
That’s how LifeLog was born: not as “yet another blog”, but as the ecosystem’s logbook.
The tools I use on this project live in the skills store — the first one is free.
Context — why another blog?
I’ve had blogs before. WordPress, Ghost, Medium — all of them died after the third post. The difference now was the motive: I didn’t want to write for an audience, I wanted to write for my future self.
The brief was simple: a devlog documenting the journey of building the Samuel ecosystem. No SEO, no analytics, no monetization. Just the story.
The stack was obvious: Astro — native SSG, MDX for content, zero JS in the build. Tailwind for styling, TypeScript for sanity. And the simplest possible deploy: Vercel, push to main, done.
The first commit was on July 4:
b5152e8 Initial commit: LifeLog Astro site com Playwright E2E tests
The name “LifeLog” came from life + log — the record of a digital ecosystem’s life. The first version was functional but Spartan: a post timeline, a simple light/dark theme, and 3 E2E tests.
The struggle — 4 themes in 2 weeks
What looked like a simple project (table + posts) turned into a UI/UX lab. In the first 10 days, I rewrote the theme system 4 times:
- CSS custom properties — simple, but no transition between themes
- Clip-path circular reveal — pretty, but stuttered on Chromium
- Native VT crossfade — smooth, but broke on mobile
- View Transition API + clip-path fallback — finally stable
Each iteration was a battle against stutter — that micro-freeze nobody asks for but everyone feels. Chromium’s View Transition API (VT) was the way out, but it needed animation:none on new(root) to avoid conflicting with the crossfade:
/* After 4 iterations — the theme that finally worked */
::view-transition-old(root),
::view-transition-new(root) {
animation: none;
mix-blend-mode: normal;
}
::view-transition-new(root) {
z-index: 100;
}
Alongside the theme came search. I started with Fuse.js — 7KB minified, it worked. But when the post count passed 30, the bundle started to weigh. I replaced it with a JSON index embedded in the HTML with word-boundary matching. Zero dependencies, zero extra JS in the build.
// The search index — embedded in the HTML as JSON, no Fuse.js
const index = JSON.parse(
document.getElementById('search-index')?.textContent || '[]'
);
// word-boundary match: "/b" + query + "/b"
And to complete the trio of problems: the tags. Tailwind 4 with color-mix and light/dark themes required opacity hacks that broke in one of the themes. The solution came with color-mix(in srgb, var(--color-accent) 12%, transparent) — which works in both themes without needing a fallback.
Resolution — when I killed the bot
On July 19, the blog already had 20+ posts. But something was wrong: the posts were reports, not stories. I was automating writing with a daily cron that aggregated commits and generated changelogs.
That’s when I realized: nobody wants to read a changelog. Not even me.
On July 24, I killed the auto-post cron and published the post explaining the turn: “From auto-post to narrative — why I killed the daily cron”.
From then on, each post became a chapter: Setup → Conflict → Resolution. Real code from the repository, extracted with git show, nothing from memory. And the weekly schedule became: one project per day, one story per post.
In late July, LifeLog also gained Open Design — a formal design system with DESIGN.md (12KB), CSS tokens, and craft rules. Every color palette, every component, every transition documented in a design contract that the site itself consumes.
Metrics
| Metric | 04/07 (birth) | 31/07 (today) |
|---|---|---|
| Commits | 1 | 158 |
| Posts | 3 | 50 (+50 EN) |
| E2E specs | 1 | 4 |
| Components | 3 | 8 |
| Color palettes | 1 | 6 |
| AI covers | 0 | 59 |
| Theme iterations | 1 | 4 |
| Build time | ~30s | ~15s |
Takeaways
The biggest lesson wasn’t technical — it was about what’s worth automating. Automate deploy? Yes. Automate tests? Yes. Automate post writing? No — because then you lose the story.
LifeLog taught me that documenting work isn’t about reporting what was done, but about telling why it was done that way.
Related entries
Same project, same ground — the stories connect.
Next step
If this was useful, there is more: I write about the same ground in the other projects, and the store holds the tools I use here.
Done — I will send the next article when it is out.
No spam, one-click unsubscribe. I am also here: Skills store · Portfolio