What if your RPG rules mostly stayed out of the way and just stirred the drama? I’ve been reading QuestWorlds, Chaosium’s generic story engine, and wrote up my thoughts. blog.monkeyx.games/2026/03/06/q...
What if your RPG rules mostly stayed out of the way and just stirred the drama? I’ve been reading QuestWorlds, Chaosium’s generic story engine, and wrote up my thoughts. blog.monkeyx.games/2026/03/06/q...
Et deuxième talk de la soirée, avec Yannick Sébastia qui nous parle de Service Mesh et Zero Trust
Salle comble pour votre soirée de mars avec Lorens qui nous parle de vi...
On verra si, en plus d'apprendre plein de choses, on arrive à en sortir à la fin du talk... 😱
Rappel : FinistDevs, jeudi 5 mars, 18h30 à Zenika Brest
Au programme :
• Lorens Kockum - “vi a 50 ans” (astuces, features… et comment en sortir)
• Yannick Sébastia - “Service Mesh & Zero Trust” (OpenShift Service Mesh 3.2, Istio Ambient…)
Infos et inscription : www.meetup.com/finistdevs/e...
Here’s the kickoff post: lostinbrittany.dev/en/why-im-le...
If you’re also learning a language “despite” agents, I’m curious which one... and why.
And I’m going to document the process: the projects, what the compiler taught me, what the agent got wrong, and the review rules I keep reusing (“cheap correctness first”, etc.).
So I’m learning Rust with a different focus than 5 years ago: reading & reviewing first, writing second.
Semantics > syntax. Failure modes > trivia.
Rust is a great example: “it compiles” doesn’t mean “it’s good”. Agents can produce clone soup, weird lifetimes, async-by-reflex, brittle error handling, dependency sprawl… all perfectly compilable.
If an assistant can generate 200 lines in seconds, the bottleneck isn’t typing. It’s judgment: is this correct? safe? over-engineered? will we regret this dependency in 6 months?
I did a basic Rust intro a couple years ago, then… never really used it. In 2026 I’m coming back to Rust on purpose, because agents exist.
QuestWorlds releases in hardcover this week!
Who is excited to add this beautiful tome to their collection?
My sons playing Tablut
We played tonight at the table. My two sons arguing over tactics with a handful of stones.
And for some moments I almost forgot grief.
en.wikipedia.org/wiki/Tablut
I drew the grid by hand and walked to the beach for pieces. Pebbles for one side, rougher stones for the other, a shell for the king.
There’s something steady about doing small, precise things.
Tablut is an old Scandinavian strategy game.
Asymmetric. One king trying to escape.
The other side closing in.
Simple rules. No randomness.
A homemade Tablet board and pieces
The last couple of weeks have been harder than usual.
This weekend I built a Tablut board.
I’d been playing it on my phone, but that didn’t feel like enough.
Que ce soit du Mans ou de Tours, l'étiquette monte aussi qu'ils savent chercher la qualité des produits à la source, avec du porc espagnol 🤩
@lostinbrittany.bsky.social vient nous parler MCP, choix et pattern.
Deuxième jour de @touraine.tech, celui des fouées 😋
On commence avec @lostinbrittany.bsky.social qui nous parle de « Serveurs MCP au-delà des bases : bonnes pratiques, choix de conception et leurs conséquences »
Ça va causer IA 🤖 un peu 😉
[𝗠𝗔𝗖𝗜] Et si l’IA n’avait plus besoin du cloud pour fonctionner ?
LLM sur smartphone, agents open source, automatisation et jQuery 4.0 : tour d’horizon des signaux faibles et forts de la tech.
Avec Julie Pouny @wanoo.me @davidcaussinus.bsky.social @lostinbrittany.bsky.social
youtu.be/Mqo5XqvtMW4
A deux doigts de me reintéresser a PHP grace a @lostinbrittany.bsky.social … non j’déconne mais super talk qui fait un bon refresh sur l’evolution du langage et l’écosystème.
3/ Slides: lostinbrittany.dev/talks/2026/2...
6-part blog series going deeper: lostinbrittany.dev/en/rewriting...
Merci au Finist'AI Club! 🙏
2/ Short answer: we're going from co-pilot to conductor. Orchestrating AI agents, designing context, reviewing everything. The vibe coders who skip understanding are learning the hard way why that matters.
The future of dev isn't less human. It's differently human.
1/ Really enjoyed presenting "Rewriting the Role: Developers in the Age of LLMs" at Finist'AI Club in Brest last night!
The gist: we've survived every "developers are doomed" wave so far. LLMs are the latest one, and the interesting question isn't "will we survive" but "how is the craft changing?"
Ça sera un plaisir!
[𝗞𝗘𝗬𝗖𝗟𝗢𝗔𝗞] Keycloak as a Service continue d’évoluer sur Clever, en partenariat avec Please Open It.
Clusterisation, console dédiée, nouveaux request flows : dernières nouveautés en live le 26/02 à 13h avec @lostinbrittany.bsky.social et @mathieupassenaud.bsky.social
youtube.com/live/Jo-Njxs...
6/ Slides are up as usual:
noti.st/lostinbrittany/EqWMPw/mcp-servers-beyond-101-good-practices-design-choices-and-their-consequences
or in:
lostinbrittany.dev/talks/2026/2...
Thanks @jfokus.se for having me!
5/ My favorite framing from the talk:
MCP in 2026 ≈ REST APIs in 2010
Same forces (domain, trust, ownership separation), but the margin for error is smaller when your caller is a non-deterministic LLM.
4/ - Operating MCP beyond one server: composition patterns, gateways, contracts and versioning, safety guardrails with risk-tiered tools. This is where the REST-to-microservices lessons really apply.
3/
- Production patterns from day one: security, input validation, testing (including LLM evaluation and prompt injection resistance), observability