gabrielstuff

🚲🔓 Un bon antivol vs un bout de plastique : pince contre meuleuse Petit rappel de la police Londonienne qui vaut pour ici à Paris ou ailleurs

@Technologi82698 Happy to help

@gabrielstuff Exact c'est ce que j'essaie de faire ! Mais étant dans le TGV avec la connexion ça va être un peu compliqué 😅

Comment vote votre rue depuis 25 ans à Paris ? 903 bureaux de vote parisiens et 44 scrutins. Tout est en ligne ! elections-paris.org/?v=new

@lossfunk @ShriKaranHanda Did you tried the latest auto-research ? Looks like a great topic.

@ShriKaranHanda After the paper was done, we tried modern agentic tools like claude code, gave them tools and instructed them to explore/learn We found it actually wrote something like this by itself (without instructing) Stay tuned for this update.

🚨 Shocking: Frontier LLMs score 85-95% on standard coding benchmarks. We gave them equivalent problems in languages they couldn't have memorized. They collapsed to 0-11%. Presenting EsoLang-Bench. Accepted to the Logical Reasoning and ICBINB workshops at ICLR 2026 🧵

kimi is here. workers ai is officially in the big model serving game blog.cloudflare.com/workers-ai-lar…

@cgtwts Very nice but really such at design integration task for now

MiniMax just dropped M2.7: > self-evolving model built for recursive improvement > 56.2 on SWE-Pro > beats gemini 3.1, nearly matches opus 4.6, sonnet 4.6, and gpt 5.4 on coding benchmarks > strong coding performance across real-world tasks > extremely low pricing vs opus/gpt

@opencode So far good result and fast but really suck at design even with the discover image official MCP. I’d like to make Gemini 3.1 pro design and M2.7 do the plumbering. Any way to standardise or always process this way ?

MiniMax M2.7 available in Go - Better at complex tasks over M2.5 - Fast - give it a plan and it runs with it - Self-evolution - do a task > check results > fix mistakes > try again

bunny.net x UpCloud 🚀 We’re joining forces to create a unified European sovereign cloud + edge ecosystem, combining high-performance cloud infrastructure with our global edge network. The integrated solution is currently in development and will launch in summer 2026.👇

@DFintelligence Elles peuvent être entièrement synthétisées, par exemple

Bon sinon, tous les journalistes qui interviewent Yann Le Cun ou écrivent des articles sur lui, c’est quand que vous allez poser la vraie bonne question ???????????????????? ELLES VONT SORTIR D’OÙ, LES DONNÉES D’ENTRAÎNEMENT ????????????????????????????? Parce que c’est bien de parler de rupture avec les LLM, mais si les LLM ont marché aussi bien, aussi vite, c’est surtout parce que la data était disponible en masse. Et malgré ça, sa collecte coûte déjà des dizaines de millions de dollars. DONC COMMENT ILS VONT FAIRE ??????????????????????????

@zeeg We’ve been building the same core block since day one. It has never been easier to go from 0 to 1. In my experience, defining clear rules, strong linting, and a consistent way of thinking and building is the way.Complexity is where real value emerges for users. It always has been

im fully convinced that LLMs are not an actual net productivity boost (today) they remove the barrier to get started, but they create increasingly complex software which does not appear to be maintainable so far, in my situations, they appear to slow down long term velocity

For weeks we were told Parisians would punish the city’s street transformation. The election results tell a different story. Pluralistic ignorance: when many quietly support something but assume others don't. Safer and greener public space may be far more popular than we think.

Apropos of nothing in particular, did you know the Cloudflare Workers runtime is open source? github.com/cloudflare/wor…

The web was amazing prior to the reactification of everything 😄. Hotwire's Turbo framework is based on Chris Wanstrath's original Pjax concept (that powered GitHub in those early years). This kind of speed is still possible!

@hugorcd Wow ! Looks great

evlog now has a filesystem adapter + new AI skills The FS adapter stores your logs locally. The skill lets agents read them. $ npx skills add hugorcd/evlog Then: /analyze-logs why did this request fail? Agents can now understand what's actually happening in your app. Production issues, script flows, debug traces... everything. Check: dub.sh/evlog-fs

@omartineau Ah oui, j’ai mixé avec Codex seulement. Ça ne devrait pas tarder avec Claude web, on devrait bientôt voir Claude Code agent débarquer

@gabrielstuff Justement il n'y a pas d'intégration native de Linear avec Claude. Ca passe par Cyrus. Et personnellement je n'aime pas trop que mon code passe par des plateformes que je ne maitrise pas.

Marcel prend forme. C'est un orchestrateur entre Linear et Claude Code

@jasonkneen @Huggiapp @aisdk OpenTUI is great but Tables support was not that great. Did it get better recently ?

Introducing "OpenTUI x AISDK" by @Huggiapp -- an @aisdk OpenTUI interface for gorgeous terminals in React + TypeScript. Primitive based TUI where everything you see is an extension that can be replaced/customised and hot reload. Create extensions to add / @ commands, create dialogs, providers, features, replace anything from a runtime core extension manager (also replaceable) and comes with skills to do all that. Write an @aisdk chat UI for web, then run the same code in a terminal. Coming soon, MIT license.

Who’s building today?

@thomasgauvin Want please 🙏

more email service private beta invites going out, keep an eye on your inbox 👀

@remicorson Quelles différences avec mesdroitssociaux.gouv.fr/votre-simulate… ?

aidesfrance.com a saturé hier soir, vous étiez tellement nombreux ! Si vous n'avez pas pu y accéder, c'est le moment : - La simulation est 100% gratuite - +70 aides analysées en quelques minutes - Un rapport détaillé disponible pour 4,99€ Découvrez vos droits sur aidesfrance.com

@joshpuckett @morganknutson @financialvice Ok simplifying then

@gabrielstuff @morganknutson @financialvice Thanks for the Vue PR! 🙌 I’ll take a look and merge later today. I think that problem is best solved with improved readme/docs, e.g. should have something like this now that we have a bunch of framework support

DialKit 1.1.0 brings: - Svelte support for all your Svickos - Ability to render panel inline vs popover (thank you @financialvice!) Enjoy!