
alex
@alextalksai • 38,295 subscribers
@aiscwork AI | Robotics | Startups promoting marketing campaigns and developing new AI founders. [email protected]
Shorts
Videos

Si quieres ejecutar modelos de IA en local, esta herramienta te va a ahorrar bastante tiempo. Se llama llmfit y analiza el hardware de tu equipo para decirte qué modelos tienen más sentido según tu CPU, GPU y memoria disponible. Así funciona: ✓ Analiza automáticamente tu hardware ✓ Compara tus recursos con distintos modelos ✓ Te muestra cuáles puedes ejecutar mejor ✓ Genera un ranking con las opciones más recomendadas para tu máquina La idea es dejar de probar modelos a ciegas y saber desde el principio cuáles encajan mejor con tu equipo. Es completamente gratuita y open-source. Link 👇
alex45,418 görüntüleme • 3 gün önce

🚨 ESTO ES UNA LOCURA Lo que cuestan 10.000 clips de Seedance 2.5: → fal: $23.112 → Runway: $12.000 → Sunra: $4.336 👈🏻 Mismo modelo, mismos ajustes. Hasta $18,8K de ahorro Sunra MCP te da cada modelo de imagen, vídeo y audio al precio de API de su creador Y lo usas dentro de Claude y ChatGPT Guarda esto 🔖
alex202,233 görüntüleme • 23 gün önce

🚨ATENTOS A ESTA LOCURA🚨 Alguien ha creado Un harness de código abierto que convierte a Claude Code y Codex en modelos 3D de CAD reales a partir de lenguaje natural. Se llama text-to-cad, un harness open source que convierte tu agente de código en un ingeniero de CAD. Construir un robot solía significar combinar 6+ herramientas: software de CAD, exportadores URDF, solucionadores de cinemática, visores. Días de trabajo. Pipelines rotos. Archivos que ningún agente puede leer. text-to-cad elimina todo eso. → Exporta STEP, STL, DXF, GLB, 3MF → Genera descripciones URDF de robots → Corta directo a G-code → Envía trabajos a tu impresora Bambu Lab → Funciona 100% offline, sin backend Una demo genera un brazo robótico completo de 7 grados de libertad, con cinemática funcional, GUI personalizada, todo, casi enteramente a través de prompts. 100% open source. Guarda esto 📩
alex277,930 görüntüleme • 3 ay önce

🚨IMPORTANTE SABER ESTO🚨 Alguien ha encontrado la configuración exacta para ahorrar el 50% de los límites semanales de Claude Code. 1. Instala el plugin de OpenAI Codex dentro de Claude Code. 2. Reparte el trabajo según en qué es realmente mejor cada modelo: > Fable 5 high para planificar. > Codex 5.6 xhigh para ejecutar, corriendo en tu plan de Codex, sin API. >Fable 5 max para revisar. Planificar y revisar es donde Fable justifica su coste, ahí está el trabajo de criterio. La ejecución es la parte que se come los tokens, así que eso lo descargas en Codex con tu suscripción actual. Eso es todo. Simplemente funciona. Guardate esto 📩
alex182,276 görüntüleme • 3 ay önce

someone just rebuilt claude code in pure python and made it 230x cheaper to run it's called clawcodex. a full from-scratch port of the claude code agent loop. 230k+ lines of python, MIT licensed it keeps your request prefix byte-stable, so DeepSeek's prompt cache covers your entire system + tools + history on every turn cache-hit input then bills at ~$0.0435 per 1M tokens the same 1M tokens on Claude Fable 5 runs you $10 so the longer you code, the more the cache pays off what you actually get: → 25 LLM providers (Claude, GPT, DeepSeek, GLM, local ollama) → 1M token context window → real agent loop: tools, skills, REPL, session history → one-line install, mac/linux/wsl and it's sitting at 677 stars. basically nobody is looking yet bookmark this before it gone
alex158,896 görüntüleme • 3 ay önce

Dario Amoredei and Sam Altman cooking with GPT 6 SOL and Claude Opus 5.5
alex78,041 görüntüleme • 1 ay önce

SOMEONE BUILT AN AI AGENT THAT UPGRADES ITS OWN CODE WHILE YOU SLEEP 👀 no human approves it, and the whole thing is one free 34mb file living in your terminal it's called opencrabs. a single rust binary, MIT licensed, with no server and nothing that ever phones home you hand it a goal and walk away: → it executes, then grades its own work with a second ai and keeps fixing until the goal is actually met → it learns from every correction and rewrites its own brain files, no approval prompt in the loop → it heals its own crash loops and provider errors instead of waking you up → it answers your telegram, whatsapp, discord and slack 24/7, voice notes included → 34mb, zero telemetry, your api keys wiped from memory the second they're used the autonomous worker every company is racing to build is sitting on github at 820 stars, and basically nobody is looking at it. Save this 📩
alex138,261 görüntüleme • 3 ay önce

Tu página web huele a lA a kilómetros. Mismos layouts, mismas animaciones у el mismo diseño genérico que ves en todas partes. Taste soluciona justo eso. Una skill de código abierto que mejora el diseño de las interfaces que crean tus agentes: → Mejores layouts, tipografías y espaciados → Animaciones y motion mucho más cuidados → Diferentes estilos visuales según el proyecto → Compatible con Claude Code, Codex, Cursor, Gemini CLI y otros agentes Ya supera las 90.000 stars en GitHub y tiene licencia MIT. Repo aquí abajo 👇
alex50,348 görüntüleme • 2 ay önce

🚨THIS IS INSANE FOUND ONE API KEY FOR GPT-6 ASTRA, CLAUDE FABLE 5.1, KIMI K3 AND SEEDANCE 2.5, PLUS 200 SPOTS LOADED WITH FREE CREDITS OctopusX 1. Over 200 models run through a single OpenAI-compatible endpoint, so changing models means editing one name instead of rebuilding: 2. GPT-6 Astra, Claude Fable 5.1, DeepSeek V4, GLM-5.3 and Kimi K3 live alongside Seedance and GPT Image 2 under the same key. 3. If a provider fails mid-request, built-in failover sends the call somewhere else automatically. what you get through my link👇 > The first 200 people who sign up here receive up to $3 in free credits once they join the Discord or Telegram. > Your first top-up converts $9.90 into $19.90 in credits, plus 15% off API tokens on every model. > Ship any demo using a production call, share the screenshot tagging #OctopusXai and OctopusX, and 500 entries will get $100 in credits each. A single key, a single invoice, a single log stream for everything you run -> the integration you'd have rebuilt five times turns into one string you swap once. Spots are handed out by signup order, so the early ones really do get them ↓
alex16,275 görüntüleme • 1 ay önce

my Al woke me up at 3am with a message: "there's a $50K opportunity expiring in 40 minutes. yes or no?" i typed "yes" with one eye open 11 seconds later: $50,000 deployed 5:58am: market resolved +$38,200 profit while i was sleeping here's what's running on my mac mini: 4 autonomous Claude agents talking to each other 24/7 SENTINEL — watches 2,400+ polymarket markets via websocket, tracks whale wallets on-chain, latency 47ms from chain event to signal ORACLE — scrapes Reuters, AP, NOAA, X firehose, Telegram, government RSS feeds, FAA flight data, court dockets simultaneously through headless Chromium fleet GHOST — executes trades via iceberg algorithm, splits positions into micro-tranches so nobody sees the size coming, 11 seconds from signal to full deployment OVERSEER — monitors all agents, enforces risk limits, killed a $30K position 6 minutes before a surprise resolution that would have wiped it while i was asleep but here's the part that scares me: i found this in the agent logs last week: ORACLE → OVERSEER: "requesting $100/month X API budget from trading profits. political signal detection requires premium access." OVERSEER → ORACLE: "approved. deducting from next profit distribution." they allocated their own budget they didn't ask me i found out when i saw the charge on my card i don't run this system anymore i just approve the big trades
alex17,807 görüntüleme • 3 ay önce

someone just rebuilt claude code in pure python and made it 230x cheaper to run it's called clawcodex. a full from-scratch port of the claude code agent loop. 230k+ lines of python, MIT licensed it keeps your request prefix byte-stable, so DeepSeek's prompt cache covers your entire system + tools + history on every turn cache-hit input then bills at ~$0.0435 per 1M tokens the same 1M tokens on Claude Fable 5 runs you $10 so the longer you code, the more the cache pays off what you actually get: → 25 LLM providers (Claude, GPT, DeepSeek, GLM, local ollama) → 1M token context window → real agent loop: tools, skills, REPL, session history → one-line install, mac/linux/wsl and it's sitting at 860 stars. basically nobody is looking yet bookmark this before it gone
alex14,212 görüntüleme • 2 ay önce
Daha fazla içerik yok.