# The blog is back!

> The blog is back after a long pause, fully rebuilt on a new stack. Now written with a well-scoped LLM, it should hold an honest cadence.

- URL : https://blog.xreveillon.eu/en-US/blog/the-blog-is-back
- Date : 2026-07-19T00:00:00.000Z
- Langue : en-US
- Tags : meta, llm

---

The blog is back, after a long pause. No suspense: this isn't a facelift, it's a full rebuild. New stack, new theme, new intentions — and, more importantly, a new workflow. The goal of this meta post is to stay short: explain what changed, why now, and what's coming. Long-form pieces will follow in the next few weeks.

## A rebuild, not a facelift

The site is rebuilt on a new foundation. A table says it better than a paragraph:

| | Before | Now |
|---|---|---|
| SSG | Hugo + PaperMod | Astro v7 |
| Typing | — | strict TypeScript |
| Styling | PaperMod theme CSS | Tailwind v4 (`@theme`) |
| Theme | PaperMod | "Terminal" (mobile-first, dark/light, self-hosted JetBrains Mono, WCAG AA) |
| i18n | FR/EN | FR/EN kept (symmetric paths) |
| Comments | Comentario | Comentario, **merged FR/EN** via `translationKey` |
| Hosting | Netlify | Netlify (Git auto-deploy) |

Every line maps to a real problem in the old site: missing types, a theme that was hard to extend, comments split in two between FR and EN. The foundation is new, not just repainted — and a lot more maintainable.

## Why now? LLMs as leverage

Writing was the bottleneck. I had (and still have) a backlog of ideas in reserve: notes, drafts, topics on hold. But turning an idea into a clean bilingual post used to take weeks — sometimes forever. The follow-through stayed rare.

With a well-scoped LLM, that math changes. Producing a first bilingual draft goes from weeks to hours. The writing mechanics are no longer the bottleneck — so the cadence will follow. That's the main reason I'm reviving the blog now, and not last year.

Let me be clear, though: the LLM doesn't do everything. Framing is still the hard work. Choosing what to write about, why, in what order, in what tone, with what judgment — that stays on me. The LLM executes; it doesn't decide. The gap between "having an idea" and "publishing it" got smaller. It didn't vanish.

## What I believe about LLMs and writing

I'm saying this as a conviction, not a trend: the text written by an LLM mirrors whoever drives it. Given raw, with no brief, no plan, no review, it produces generic text — readable, flat, with no angle. Driven by someone who knows what they want to say, it becomes a real accelerator.

What's interesting is that the tool lowers the barrier to writing. Writing something readable has never been more accessible. Anyone can produce a decent post, in a language that isn't theirs, on a topic they struggle to put into words.

And that's where the usual objection comes up: "everyone will write the same thing, quality will collapse." I don't buy it. Quality doesn't collapse — it shifts. The barrier is no longer writing; it's framing. And framing is exactly what already separated a good post from a lukewarm one before LLMs existed.

This isn't magic. It's closer to training wheels: the small wheels stabilize, but you pedal and you steer.

If framing is the real work, then the posts that follow are worth reading — not because they're "AI-assisted," but because the time saved on mechanics gets reinvested in substance.

## No tracker, no cookie

This blog has no analytics. No tracker, no cookie, no GDPR banner asking you to accept fifty partners. It's a pure static site, served by Netlify.

That's a choice, not an oversight. It's consistent with the rest of the site and with my topics: on-prem, open-source, sovereignty. If I want to know whether a post resonates, I'll read the comments. That's enough.

## What's coming

What's in the pipeline, in no particular order:

- **Pragmatic SSL/TLS** — a long-form piece that's been a long time coming, on PKI, short-lived certificates, and how to set up a chain of trust without overcomplicating things.
- **"A website has three audiences"** — a reflection post on who you write for when you publish online (readers, machines, yourself).
- **Dagster** — the follow-up to the [data-platform](/en-US/blog/data-platform) series, this time on pipeline orchestration.
- And more ideas in the [backlog](/en-US/blog/backlog) — where you can also suggest topics in the comments.

The point isn't to ship everything at once. It's to hold an honest cadence, in both FR and EN.

Welcome — or welcome back — to the blog. The [RSS feed](/en-US/rss.xml) is open, and so are comments, merged between FR and EN — write in whichever language you want. See you soon for long-form content. If a topic matters to you, the [backlog](/en-US/blog/backlog) is open to suggestions.