Mossad Inside

@PUMAPARDO_ @GordoDan_ @PabloAbrazo_ Querrás decir 120p 2Fps

@GordoDan_ @PabloAbrazo_ una mamushca de gordos stremeando con cada vez menos calidad hasta llegar a 240p

🚨🚨MINUTO A MINUTO DE LA SITUACIÓN EN USA POR EL ATAQUE A TRUMP 🚨🚨 🚨🚨RUCKAUF🚨🚨

Workflow builder

@prin_vlr x.com/i/status/19868…

Che es increíble lo desastroso que es el Banco Galicia. Me volvieron a vaciar la cuenta usando mi tarjeta de débito NUEVA que me habían enviado hace menos de dos semanas por el último “hackeo”. Otra vez tengo más de 20 compras seguidas que no hice y encima las últimas tienen fecha futura??? Su seguridad es un chiste @BancoGalicia

@tendenciaytuits @JMilei Sacado de IA, falta la novela de frutas

“Peter Thiel” Peter Thiel acaba de comprar una mansión de 12 millones de dólares en Palermo Chico, Buenos Aires 🇦🇷 No Miami. No Mónaco. No Londres. Buenos Aires. 1.600 m² en la zona más exclusiva y protegida, donde viven embajadas y la vieja aristocracia. La cuenta es brutal: solo 700 dólares por pie cuadrado. En Manhattan vale 4k-8k, en Mayfair 3.5k-5k. Thiel pagó entre el 15% y 20% del precio real de un trofeo similar. Una ganga sospechosa 🧐 Patrón clásico de Thiel: En 2011 sacó ciudadanía en Nueva Zelanda 🗺️ En 2021 se mudó a Miami antes que nadie 🏠 Ahora aterriza fuerte en Argentina con 36 millones en propiedades en 16 meses 💰 Su red siempre lo sigue 18 meses después. Socios, empresas, capital… todo se mueve. Thiel no está comprando una casa. Está comprando una opción call física sobre el próximo gran arbitraje geopolítico 😏 Cuando Peter Thiel empieza a comprar mansiones en tu país… ya no es turismo. Es una señal. ¿Visionario o buitre oportunista? ⏳

kinda obsessed with stone shade.

parááàáàááá

The Local LLM Cheat Sheet for Your 128GB RAM Device ✅16GB ✅32GB ✅64GB ✅128GB 🔜256GB??? At the 128GB tier, you can run serious all-rounders, agent/coding models, math specialists, vision models, and even long-context research models. The big takeaway is that there is no single “best” model here. At 128GB, model choice becomes highly use-case dependent. General Purpose & All-Rounders Qwen3.5-397B-A17B / GGUF / UD-IQ1_M The strongest overall all-rounder on 128GB. Best for general assistant work, hardest reasoning, and complex agents. This is the headline pick if you want one local model that can cover the broadest range of serious workflows. Mistral Large 2411 / GGUF / Q4_K_M Best writing and multilingual all-rounder. Strong for creative writing, business prose, multilingual tasks, function calling, and long-form output. Not the top benchmark pick on the sheet, but excellent when language quality, tone, and writing polish matter. Coding / Agent / Tool-Use Kings MiniMax-M2.7 / GGUF / UD-IQ2_XXS Best agent-loop and tool-use pick. Strong for tool calling, multi-step coding, autonomous tasks, and agentic workflows. This is one of the clearest choices if you are building local agents or want a model optimized around tool use. GLM-4.5-Air / GGUF / Q6_K Best coding and SWE-focused option. Strong for repo work, SWE tasks, agent tool loops, code edits, and practical engineering workflows. Pick this when you want a strong balance of coding ability, reasoning, and tool use on 128GB. Math & Competition Reasoning Qwen3-235B-A22B-Thinking-2507 / GGUF / UD-Q2_K_XL Best deep math and structured reasoning model. Strong for competition math, deliberate analysis, structured problem solving, and deep logic. This is not the everyday general chat pick, but it is one of the strongest options when the work is explicitly reasoning-heavy. gpt-oss-120b / GGUF / Q8_0 Math and efficiency veteran. MXFP4-native weights make Q8_0 effectively full precision at a very efficient size. It is an older model from August 2025, but it is still extremely strong for math, olympiad-style tasks, hard reasoning, and efficient high-quality local inference. Vision & Multimodal Qwen3-VL-235B-A22B-Thinking / GGUF / UD-IQ3_XXS Best vision and multimodal specialist. Use it for image understanding, document analysis, OCR-style workflows, UI analysis, charts, diagrams, and multimodal agents. If you want a serious local vision model in the 128GB tier, this is the one to look at. Long-Context / Research / Fast Thinkers Llama-4-Scout 17Bx16E / GGUF / Q6_K Best long-context research specialist. The reason to pick this is the 10M native context window. Use it for massive document sets, full-codebase Q&A, archives, research corpora, and workflows where context length matters more than raw benchmark quality. Qwen3-Next-80B-A3B-Thinking / GGUF / Q6_K Best fast thinker. Strong for fast reasoning, interactive math, agent planning, and responsive long-context use. Pick this when you want something that feels quick and useful in live workflows, not necessarily the deepest possible reasoning model. Let me know which models you are running on 128GB, or which memory tier, hardware profile, or model category you want next.

damn, our team is moving so fast. we're designing 10s and 10s of screens/flows across multiple projects on a daily basis! best part is, the quality is still world-class 😁

Minimal UI. Maximum obsession.

Clean Onboarding Progress UI

My most favourite recent work, Did i cook?

User info dropdown

first-run experience for conversational demo agent

¿?

Así están los rusos en X explicándole a los comunistas de café porque la Unión Soviética era mala.

"Farsante" Porque el estadounidense Dante Guebel dio una entrevista y no resistió ni su propio archivo.

I think what I’ve come up with so far is clean.