DonWlad

@sama gpt-oss-codex 👀👀👀

@RappiColombia suma y sigue me fui a comprar los ingredientes las preparé me las comí y aún nada, rappi cada vez peor... Suscripción black pro no sirve para nada a cancelar... Adiós. De las 09:20 a las 11:39 por favor.

@RappiColombia es enserio!!!! Pago suscripción pro, y nada, ni una explicación ya tengo varios pedidos en el mes que llegan a 2 horas de espera y el rappi no llega. Si no van llegar es mejor que cancelen y no le hacen perder el tiempo a uno.

@barckcode Totalmente de acuerdo!!!

La IA ha conseguido lo que ningún PM ha conseguido en su vida. Que documentemos todo con detalle para darle contexto a la IA 🤣

@GabrielBoric Jajajaja claramente si boric está en eso, todo es una mierda jajjajaja

La inteligencia artificial es la mayor revolución tecnológica del último tiempo y desde Latinoamérica y el Caribe es estratégico y urgente que tengamos un rol. Hoy desde Chile lanzamos Latam-GPT, un modelo de lenguaje abierto y colaborativo, resultado de una alianza con 15 países, que permitirá incluir datos e identidad latinoamericana en nuestra interacción con la IA. Soberanía tecnológica y cultural para que nuestra región sea parte de esta conversación global. Felicitaciones al Ministerio de Ciencia, Tecnología, Conocimiento e Innovación y al Centro Nacional de Inteligencia Artificial por impulsar esta iniciativa que nos sitúa en el mapa.

@0xSero I use GLM 4.7 for the agent, Playwright MCP snapshot for code review, and local MCP with my Qwen vision model for reviewing PNG screenshots of the frontend. I have an RTX 3090. Is what you're suggesting better than that?

Let me save you hours of testing frontends. If you're ever working on a front-end, instead of writing tests, and adding puppeteer slop to your repo 1. Get an llm to write you expected.md with whatever needs to be tested 2. Copy that, go to browser 3. Open localhost with your selected app 4. Use Claude Chrome Extension or Parchi 5. Send it the expected.md prompt 6. QA engineering, there you go. Use models results and pass it back to your coding agent to fix whatever is flagged.

@0xSero Is very good, Other subscription to list a pay jajajja. I have 8 rtx3090, only need motherboard is connected all GPUs. This year make the proyect. your videos so good.

Kimi-K2.5 before and after for vllm-studio Before: After:

@0xSero Congratulations 🎉🎉

My daughter is born, no longer hypothetical

Experience Z-Image Turbo - Generate photorealistic images in just 8 steps! z-image.ai

@MonicaSaadeX Primero 2 por qué va recto y no tiene señal de tránsito, luego el 3 por viene por via sin señal de tránsito que diga lo contrario y por último 2 tiene una señal de ceder el paso hasta que tenga el espacio para avanzar por lo tanto pierde preferencia.

🤔👇

@0xsero Excelent the new era to programming from ai is awsome. from my side integrated info to mudblazor docs to RAG system in to kilocode and frontend is a game. Your info is very util. Regards from Medellin colombia.

I started off my career in tech as a JS front-end dev mainly, I worked on a lot of applications and websites with React class components, and later moved on to hooks, then Next, etc.. After about 2-3 years of building frontends I had gotten absolutely sick of dealing with UI/UX, it became so boring and monotonous The work is so detail oriented, endless, and not well respected by my peers. So I spent a year doing anything I can to sell myself as a back-end dev, I built bots, servers, scrapers, on-chain products, automation, etc.. I learned some Rust, Python, Solidity, etc.. anything to do something more fun, and engaging, so I don't have to spend hours debugging stupid Next/React rendering loops, etc.. I thought I would never touch a front-end again, at least not willingly. This weekend I got a nice offer from a friend to help them build something cool, so I jumped at it. Just a weekend, no need to stress about it. So I fired up Cursor with Composer-1 and Figma MCP. I can honestly say I have not felt this excited about working since my first 6 months coding. This thing is an absolute power-house, it's sooooo good. I have been blasting pages faster than I can finish my 1 hour playlist. It all works perfectly. Sometimes I bring in GPT-5-Codex-high and Claude-Sonnet-4.5 to do more complex fixes for layers and transitions. This is a new paradigm of work, whoever adopts this can scale to building dozens of functional applications a month. I remember when I was a junior/mid-level getting paid 3-6k USD a month to build MAYBE 2 pages.

@nisten Hahaha I'll keep my rtx 3090 ti x 6! 😎

Nvidia DGX 128gb $4000 mini pc benchmarks are in.... Barely gets 11tps on gpt-oss-120b fp4... 6tps on qwen3-32b-fp8 even with sglang optimizations Yeah this is a flop, buy an actual gpu.

@MovistarCo que paso hoy con las 900mbps? En Medellín velocidades que me hacen recordar cuando tenía dial-up 56kbps pero por fibra ;)

@MudBlazor It’d be awesome if you provided an MCP/retriever for the docs (indexed in a vector DB like Qdrant). Agents could query the docs in real time—better bug fixes, citeable snippets and much more effective “vibecoding”. Interested in a prototype collaboration? 🔧✨

we have signed a deal for an additional 4.5 gigawatts of capacity with oracle as part of stargate. easy to throw around numbers, but this is a _gigantic_ infrastructure project. some progress photos from abilene:

@damianplayer Ai

Burned out trying to stay consistent on 5+ platforms? Still paying $2K/m to an “agency”? This AI Agent creates & posts VIRAL content 24/7. – It scripts platform-specific posts – Auto-generates images – Approves via email – Posts to X, IG, FB, LinkedIn, YT++ – Archives to Drive Looking to save $2,000+ & use this content machine? Like + Comment “AI” & I’ll send it over... (must be following so I can DM) Here’s the full system 👇

@iamfakhrealam It seems that OpenAI likes Spanish, the blurry screenshot, and your video.

Q1: Find the number of triangles in the given figure? Correct is 36

DeepSeek vs OpenAI vs Qwen 🔥 I asked 8 analytical reasoning questions. Check each tool's response:

@nayibbukele Presidente Bukele, excelente observación. El overhead en IA puede reducirse optimizando infraestructura (GPUs sin ciclos desperdiciados), datos sintéticos supervisados y agentes con ajuste dinámico de prompts. Por aquí a la orden si busca potenciar a El Salvador con IA.

So, >95% of the cost of developing new AI models is purely overhead? 🤔

@MrBeast The most epic content deserves a legendary home! If I win, I’ll share the mystery with everyone who liked this comment and host a giveaway where 10 random people will get part of the prize in crypto as a thank-you! 🎉🔥 #BeastMode #BeastGames

Most liked reply on this tweet gets what’s in this mystery box! Will pick winner in 72 hours 🤪 (btw episode 6 of beast games just dropped go watch)

@MrBeast The most epic content deserves a legendary home! If I win, I’ll share the mystery with everyone who liked this comment and host a giveaway where 10 random people will get part of the prize in crypto as a thank-you! 🎉🔥 #BeastMode #BeastGames