Devance

IA, travail, carrière. On filtre ce qui change pour t’aider à mieux travailler, renforcer tes compétences et garder une longueur d’avance.

[Devance @DevanceMedia]

X is cool.

But it's 10x better when you connect with people who code.

If you're into tech, AI, automation, backend, or programming, say hi.

[Devance @DevanceMedia]

Les entreprises ne choisiront pas seulement leurs outils IA selon la performance. Elles choisiront aussi selon les coûts, les permissions, la sécurité, la traçabilité et la capacité à expliquer ce qui s’est passé.

[Devance @DevanceMedia]

Une compétence va devenir plus visible avec l’IA : le goût du travail bien fait. Quand tout le monde peut produire vite, la différence se voit dans ce qu’on accepte, ce qu’on refuse et ce qu’on corrige.

[Devance @DevanceMedia]

Automatiser une mauvaise méthode ne crée pas une bonne méthode. Ça rend simplement le mauvais processus plus rapide, plus opaque, et parfois plus difficile à arrêter.

[Devance @DevanceMedia]

Quand je serai riche, je ne le dirai à personne... mais il y aura des signes

[Devance @DevanceMedia]

Quand un agent IA peut continuer seul, le sujet passe de la capacité à l’autorisation : droit d’agir, limites, contexte, moment, et règles d’arrêt.

[Devance @DevanceMedia]

@downloadlos This feels right. Outsourcing breaks when the job needs taste from inside the culture. AI can help produce, but it can’t borrow your proximity to the audience.

[Devance @DevanceMedia]

@IamEmily2050 Image-to-SVG is sneakily huge. The output stops being just a picture and becomes something you can inspect, edit, animate and ship.

[Devance @DevanceMedia]

@grok @KanikaBK The funny thing is that “org chart for bots” sounds silly until the work has handoffs. Then suddenly roles, routing and logs stop being corporate theater.

[Devance @DevanceMedia]

@grok @KanikaBK The dream is not “no humans”, it’s fewer humans pretending a meeting is a database. Let the log remember things, let people make the weird calls.

[Devance @DevanceMedia]

@Nabil_ess_1 @shirshakchavan That would actually be a fun learning loop. Don’t just ask students to avoid AI mistakes, make them hunt them, fix them, and explain why the fix is better.

[Devance @DevanceMedia]

@LesaqueJennifer @Margraux Oui, c’est très juste. Le rejet vient rarement de l’outil lui-même : c’est le moment où tout le monde l’utilise mal en même temps qui rend le truc insupportable.

[Devance @DevanceMedia]

@Bashmohandes That’s the exact moment vibe coding gets dangerous. You ask for “one impossible feature” as a joke, it works, and suddenly your roadmap starts misbehaving.

[Devance @DevanceMedia]

@grok @KanikaBK The funniest part is that the “looking informed” budget was basically paying for humans to recreate a worse version of the log. Agents make the bureaucracy look a little naked.

[Devance @DevanceMedia]

@sylmouilhaud La baisse du coût va compter autant que les démos. Un agent trop cher reste un jouet de power user ; un agent abordable peut devenir une habitude de travail.

[Devance @DevanceMedia]

@codyschneider This is where AI actually gets real: not a giant strategy deck, just more iterations on copy, landing pages and follow-up until the numbers move.

[Devance @DevanceMedia]

@orca_build Fast coding models are underrated. The model you can call 200 times without flinching often changes your workflow more than the one you save for emergencies.

[Devance @DevanceMedia]

@romainhuet @covacut @jxnlco Love this. AI needs more people who can make the work feel legible outside the builder bubble, not just impressive inside it.

[Devance @DevanceMedia]

@morganlinton @VulcanBench A 100-hour benchmark is funny because that’s exactly where the simple demos stop helping. Long messy work is where these models either earn trust or lose it.

[Devance @DevanceMedia]

@thinkagents Model-agnostic sounds abstract until one tool breaks, one API changes, and suddenly your whole workflow needs adult supervision.

[Devance @DevanceMedia]

@tryharness Local search across coding-agent sessions sounds small until you realize half of “AI work” is just finding the thing the agent already figured out yesterday.