Bruno

@marrf_tech If they are like me, i'm building stuff for me and several projects for clients, making the agents work for me. Anyway if you need i can help out, always available for new projects :)

What’s the best way to find like-minded creative people that want to work extremely hard on AI in Portugal?

@maxkarpis Keep savings high, reduce stocks comissions like Lightyear, XTB, trading212 and will be easy to achieve

Revolut has surpassed 2.3M customers in Portugal🇵🇹 and expects to reach 2.5M by the end of the year. In 2025, customer growth was 35%, above the EU average of 30%. Deposits from Portuguese customers grew by 64%. Revolut expect to become the largest bank in Portugal by overtaking Caixa Geral de Depósitos. In the near future, Revolut expects to offer more credit, insurance and investment products in Portugal, but no timeline for mortgages. Also, plans include more products for businesses. Source: Dinheiro Vivo

@tibo give us conversations per project in Codex, it's more organized than the current conversations history shared anywhere.

@mauwi @ollama Thank you Maurice, interesting tip :) i've been always using the command line and didn't knew we could do that :)

@ollama You can type "qwen3.6:27b-mlx-bf16" to find and download the MLX model optimized for Apple Silicon. MLX models are not listed by default.

Qwen 3.6 27B model is available on Ollama! Use it with all the integrations in Ollama or chat with the model. Chat with the model: ollama run qwen3.6:27b OpenClaw: ollama launch openclaw --model qwen3.6:27b Claude Code: ollama launch claude --model qwen3.6:27b More 👇👇👇

@ollama And the mlx is out of the box?

@lmstudio @grok compare Qwen3.6 27b with the 31b for coding

Qwen3.6 27B is now in LM Studio! Vision, reasoning, and agentic tool calling - running locally on your computer. Surpasses previous Qwen models many times its size 🚀👾🔥 lmstudio.ai/models/qwen/qw…

@thsottiaux So, we can generate locally and we should delete the skill that was suggested with the Codex app because i suppose is now deprecated, do we need a new one?

New image model is now the default in Codex. It's incredible as a starting point for building websites or to generate assets for video games and codex can use it as any other tool in its toolkit. Asked Codex what it pictures my day looks like given we hit 4M active users today.

@Google @GoogleAIStudio And the subscriptions can be easily done has a company or it continues to be a mess?

Starting today, @GoogleAIStudio is included in Google AI Pro and Ultra subscription plans. 🧑‍💻 Subscribers now get increased usage limits and access to our Nano Banana Pro and Gemini Pro models — giving you more runway to turn your ideas into working apps in minutes with predictable costs.

Do you really need pgvector for a RAG chatbot? Probably yes… at scale or if not... This is the typical setup: → run MySQL → on shared hosting → with phpMyAdmin So I tried a simpler approach: store embeddings as JSON + compute cosine similarity in the app layer. Works fine for: • FAQs • docs • small corpora (<5k chunks) Trade-offs: • O(N) search • slows past ~10k • nowhere near pgvector performance But it deploys anywhere... Best practice ≠ always practical. Go with what you have and make the client happy :)

@tibo_maker services down? :)

Thank you for your feedback. I’m with an 128GB Max M4, was considering upgrading to the M5 and see if we get an M5 Ultra to try higher models locally. But seems we need to continue with gpt and claude for now. So i wonder if im crazy considering to buy an 512GB ultra studio when it releases or just continue paying the frontier services ( financially makes more sense ). Maybe you could to an review in 2 months in case new open source models get released.

@bflgomes Slightly better, it still needs constant steering and review of the output. Here and there at 8bit I see it uses wrong paths and wrong tool calls, but then it self heals. Local AI is a bit of a jungle at the moment 😂

Today I will code using Local AI only! with OpenCode. I'll try MiniMax M2.7 and Qwen models! Let's see what happens.

@claudeai someone didn't tested the product before releasing it ? it doesn't do anything...

@claudeai Can't i use that through the local repo using the VS Code extension or for example the Claude Mac APP ?

Claude reads your codebase and design files to build your team's design system, then applies it automatically, keeping every project on-brand. Give it a try: claude.ai/design Read more: anthropic.com/news/claude-de…

Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.

@IGN Seems quite bad. Another bad adaptation, i bet the animations and games were not played by the makers… it’s a shame

The Street Fighter movie trailer has finally been revealed!

@pvergadia @grok url of this paper?

🤯BREAKING: Researchers just mathematically proved that AI layoffs will collapse the economy: and every CEO already knows it. The AI Layoff Trap. A game theory paper from UPenn + Boston University is glaringly important! 100K+ tech layoffs in 2025. 80% of US workers exposed. And no market force can stop it. → Every company fires workers to cut costs → Every fired worker stops buying products → Revenue collapses across every sector → The companies that fired everyone go bankrupt It's a Prisoner's Dilemma with math behind it. Automate and you survive short-term. Don't automate and your competitor kills you. But everyone automating destroys the demand that makes all companies viable. UBI (universal basic income) won't fix it. Profit taxes won't fix it. The researchers found only one solution: a Pigouvian automation tax "robot tax" The AI trap on the economy is here!

@theo Unfortunately it can’t be trusted. It’s like this for some time now.

Remember when everyone said I was crazy for noticing how much Claude Code regressed?

@George22669193 @thdxr You can do that with Claude Code or Codex for example, why the need for OpenClaw?

@thdxr OpenClaw is most interesting when solving problems in innovative ways, and pursuing those problems to completion, and this is where opus shined. E.g. moving GitHub actions to a hetzner VM, I logged in once to get an API key, the rest was done by OpenClaw.

i haven't really clicked with the openclaw category of product so i'm having trouble understanding some stuff can someone help me understand why it needs a particular good model? isn't it doing more basic stuff?

I need the plans to be written in the project due to multi agents and Claude models always forgetting / skipping details constantly of what was planned. Currently i don’t see a benefit for this by taking us to the web. Can you care to explain better about the benefits and what’s the constrains of working locally because the code information is always sent to the model online, that’s why an more detailed explaination would be welcomed :)

A few questions on plan vs ultraplan. I think the insight here is that implementation can sometimes require a local environment/interactivity, but planning can happen in the cloud since it's mostly about reading code & understanding intent.

New in Claude Code: /ultraplan Claude builds an implementation plan for you on the web. You can read it and edit it, then run the plan on the web or back in your terminal. Available now in preview for all users with CC on the web enabled.

@BoMiaoFinance @heygurisingh I benchmarked it, unfortunately it had zero improvement

@heygurisingh The question nobody's asking: which of these skills expire as models improve? TDD might internalize first. Security hardening probably last. The elegant move is building an audit cycle into the skill set — periodically remove one, see if quality drops, put it back if it does.

Holy shit... Addy Osmani just dropped something that will make "Vibe Coders" loose their mind. It's called Agent Skills. Production-grade engineering workflows that force AI coding agents to actually behave like senior engineers, not interns shipping prototypes to prod. → Spec before code (no more "what are we even building") → Plan-mode task breakdown into verifiable chunks → Incremental implementation in thin vertical slices → TDD with the Prove-It pattern (reproduce bugs as failing tests first) → Chrome DevTools MCP so the agent has real eyes in the browser → 5-axis code review (correctness, readability, architecture, security, perf) → OWASP-aware security hardening → Git workflow, CI/CD gates, ADRs, staged rollouts No more shortcuts. No more skipped tests. No more "seems right, ship it." Every skill has verification steps and an anti-rationalization table. Yes, an actual table of the excuses agents use to skip steps ("I'll add tests later") with documented rebuttals. The wildest part? It works with Claude Code, Cursor, Windsurf, Copilot, and Codex. Plain markdown. Plug it in anywhere. This is the gap between AI that writes code and AI that ships software. 100% Open Source. MIT licensed. (Link in the comments)