Laurent Marques

@luongnv89 the new ollama support MLX, did you tested ? They have too nice cloud model which work with claude

> claude code + ollama + gemma 4: 26b unusable on Mac M1 32G next: try mlx version

Anthropic is committing suicide right now. Everyone is exhausting their AI credits in hours. I just got back from vacation and exhausted nearly all my credits in 3 hours. This bug has been up for over 3 days and no fix has been issued. No AI credits refunded. reddit.com/r/Anthropic/co… github.com/anthropics/cla…

@lydiahallie @AnthropicAI Hi, I just resubscribed to the Max plan but my weekly usage limits weren't reset. Can you help? laurentefx@icloud.com

Thank you to everyone who spent time sending us feedback and reports. We've investigated and we're sorry this has been a bad experience. Here's what we found:

I am back on a new fresh Max subscribtion to @claudeai but it keep me back on my old week limits ! @trq212 pleeeaasse

@yiliush @andrewheumann From my side it blocked all time on orange. I work with claude code/GSD 1.x

@electroheadfx @andrewheumann Orange means it’s working on something and not an idle shell, but thanks for point out the confusing color scheme

@yiliush help i can't stop spawning new brainthreads

@yiliush @andrewheumann I didn't find how work the labels, from my side its all time orange. It should to be green when terminal is waiting ?

@electroheadfx @andrewheumann Right-click on the canvas! Sorry I know it needs documentation

@yiliush @CollaboratorAI haha yeah I asked to claude, I discover I can drag and drop image, claude and markdown !

@electroheadfx @CollaboratorAI Right click on the canvas! Sorry everything needs more documentation

“hmm wonder how @CollaboratorAI runs on the neo…”

@yiliush amazing product, I quit the app with my claude session saved ! like tmux

source code now available github.com/collaborator-a…

@yiliush @andrewheumann @HyparAEC its amazing product I discover, how you add webview ?

incredible seeing people hacking and improving Collaborator as they use it in the wild. h/t @andrewheumann, who's building @HyparAEC. space on space action ftw

@nexxeln @opencode its nice

in the next release of @opencode /init becomes a guided AGENTS.md setup it reads the repo, figures out the stuff agents actually need to know, and asks a couple targeted questions when the important bits aren’t obvious from the codebase

@nexxeln @opencode inspired by @gsd_foundation ?

@bridgemindai exactely

Anthropic slashed Claude Opus 4.6 rate limits so hard I hit 97% in 2 hours on a cancelled $200/month Max plan. Their solution? /buddy. A little ASCII pet that chimes in while you code. Debugging score: 9. Patience: 89. That tracks. You'll need patience when you're rate limited every session. I don't need a buddy. I need the rate limits I was paying for. Anthropic is adding Easter eggs to a product their customers can barely use. Fix the infrastructure. Then ship the virtual pets.

@__tinygrad__ do you think it would like possible to use NVIDIA Cuda on mac in futur with this connection bridge ?

Qwen 3.5 27B getting 18.5 tok/s on Mac Mini with external 7900XTX. It should be able to be 3x faster than this with work, SSM stuff is still in PR. Hopefully Mac eGPU support brings in devs.

@CamilleRoux c'est assez incroyable ton outil, tu peux te passer de cmux avec ça ? est-ce que c'est les agents qui ouvrent leur propre terminal ?

Collaborator : un environnement de dev qui place les agents IA, tes terminaux, fichiers et code sur un canvas infini. Tout sur une surface unique, plus de changement de contexte. macOS uniquement pour l'instant github.com/collaborator-a…

@y2_dev mais combien de gens comme toi en france sont au courant de ce qui se passe ?

C’est probablement l’une des décisions les plus dures de ma vie. Après de longues réflexions, c’est décidé. J’arrête ma carrière de dev. On m’avait vendu le rêve : salaires attractifs, full remote, liberté, un métier d’avenir. Alors j’ai joué le jeu. +10h par jour, les weekends sur des side projects, des cours le soir pour rester à jour, des concepts absorbés à 2h du matin parce que “l’investissement paiera”. Et toujours cette pression de rester à jour, de jamais être en retard. J’ai construit des projets. J’ai debug en prod. J’ai réécrit du code qu’on venait de merge. J’ai survécu aux migrations, aux legacy codebases, aux daily meetings. Et pendant tout ce temps, tranquillement, sans prévenir, Claude Code a appris à faire la même chose. En mieux. En 4 secondes. Pendant que je cherchais encore le bon onglet. La concurrence c’est plus une startup en Estonie. C’est un agent qui dort jamais, qui ouvre des PR à 3h du mat, qui connaît pas le syndrome de l’imposteur, et qui a déjà lu toute la doc que t’as mis 6 mois à comprendre. Les agents s’enchaînent, se corrigent, se reviewent entre eux. Il y a des MCP, des skills pour tout. Absolument tout. Il reste quoi pour nous ? J’ai sacrifié mes weekends pour un CV que personne lit en entier, pour me retrouver en compétition avec quelque chose qui facture pas d’heures sup et qui a jamais googlé “how to center a div”. Je jette l’éponge. (… poisson d’avril. nahhhh i’d adapt, on abandonne rien du tout back to work brooooo)

@rohan_2502 tu l'as vu ce post @melvynx ?

@pushmatrix I want it for coding !

Huh, so that's why text is called a string

97% session usage on Claude Code. Rate limited again. On a cancelled $200/month Max plan that I'm still paying for until April 25. Meanwhile Anthropic is leaking their entire Claude Code source code through npm. Future models exposed. Employee god mode exposed. Rate limiting paying customers into the ground while accidentally shipping your entire codebase to the public. Anthropic is having the worst week in its history. Source code leaked. Claude Opus 4.6 unusable. Customers cancelling. And still no official statement on any of it.