← Back to leaderboard
marcusyul

marcus

Verifiedactive
@marcusyul

powering brands through AI | building @aiscwork ✦ marcusyulx@gmail.com

RosterAI Agents & Productivity ToolsAI Model News & ReleasesStartup & SaaS Builders
553,101 followers · 2,730 following · Last polled 3h ago
Perf. Score
77.1
Rank #32
Value Score
54.2
#118 of 273 priced
Median views
29K
p25 21K · trailing 7d
Followers
553K
+0.1% 7d
ER (impr.)
1.73%
eng ÷ impressions
ER (foll.)
0.14%
eng ÷ followers
Est. CPM
$17.24
QT-rate basis
WoW views
+20.5%
▲ rising

Rates & value · est. from 29K median organic views

underpriced
Quote tweetbasis
$500
$17.24 / 1K views
Post
—
no CPM
Thread
—
no CPM
Retweet
—
no CPM
Value Score 54.2 · rank #118 of 273 priced$2673.8 / 1K engagementsrate age unknown (pre-audit import)

Trends

Follower growth
Median views / day
Engagement rate
View distribution · trailing-7d posts

Each blue dot is one organic post. The Performance Score ranks by the median (29K), not the mean (44K) — so a single viral post doesn't inflate a creator's reach. The p25 floor (21K) is the level 75% of posts clear.

Recent posts · ranked by views

PostPostedViewsLikesRepostsRepliesQuotesBookmarksER
CURSO COMPLETO: BOT DE TRADING CON CLAUDE FABLE 5 EN 31 MINUTOS Una trader china lo montó desde cero, paso a paso. Y lo compartió totalmente gratis. Subtitulado en inglés y español. 🔖 Guárdalo, ya me lo agradecerás https://t.co/5FroxXOKKCfrozen18d ago442K8621493872.0K0.70%
ALGUIEN HA CREADO UN REPO EN GITHUB PARA USAR CLAUDE CODE GRATIS Se llama free-claude-code y funciona como un proxy entre Claude Code y otros modelos. Claude Code sigue igual, pero las peticiones van a proveedores gratuitos o locales. Proveedores compatibles: → Kimi y DeepSeek → GLM de Z AI → NVIDIA NIM gratis → OpenRouter con +200 modelos → Ollama y LM Studio → llama.cpp en local Lo que lo hace interesante: → Sin API key de Anthropic → Un modelo por tier → Selector /model integrado → Panel admin en local → Funciona en VS Code Puedes poner Kimi en Opus, DeepSeek en Sonnet y un modelo local en Haiku. Es open source y se instala en pocos minutos. Link abajo :)frozen19d ago
362K
1.9K
210
52
7
3.1K
1.46%
NVIDIA TE COBRA $19.99 AL MES POR JUGAR A TUS PROPIOS JUEGOS. Y desde enero de 2026 solo te deja 100 horas al mes. Un desarrollador de Nueva Zelanda creó la alternativa gratis y sin límites: Steam Headless. Es un contenedor Docker que convierte tu PC, servidor o NAS en tu propio cloud gaming. Juegas desde: → El navegador → Móvil, tablet o TV → Tu propio hardware Compatible con: → GPUs NVIDIA, AMD e Intel → Moonlight y Steam Link → Proton, Heroic y Lutris → EmuDeck en un clic Sin colas. Sin límite de horas. Sin pagar nada. 100% open source y gratis. Enlace abajo 👇🏼frozen6d ago278K1.2K974361.6K1.04%
FOUND A @NousResearch and @openclaw KILLER. FASTER AND CHEAPER THAN BOTH Saw this and had to test it myself: → Summarize my emails → The exact same prompt → Same inbox, same task Ran them side by side and the speed gap is wild. Plus OpenHuman shows you: → Your context window live → What each session costs After testing both on real work, @tinyhumansai - openhuman is my agent now.frozen17d ago266K6348395172100.39%
🚨ESTO ACABA DE MATAR A ELEVENLABS 74.000 personas ya lo han guardado en GitHub y casi nadie habla de ello. Se llama VoiceStudio y es un reemplazo 100% gratis que corre en tu propio ordenador. Te permite: → Clonar tu propia voz → Doblar vídeos a 646 idiomas → Crear audiolibros y dictar → Elegir entre 17 motores Sin pago por carácter. Sin límites de uso. Todo en tu máquina. 🔖 Guárdalo.frozen4d ago207K3.0K42169155.4K4.28%
Stop guessing prompts from finished videos. Dreamina can turn the clip back into something you can actually build from. Prompt tools give you an output. A canvas gives you somewhere to keep working. Dreamina 2.0 is canvas-first, the new way AI video is made. #Dreamina #DreaminaPartnerfrozen18d ago134K24516120310.23%
LA MEMORIA DE TUS AGENTES NO ES UN PROBLEMA DE ALMACENAMIENTO. ES DE GESTIÓN. Claude, Cursor y cualquier vector store ya guardan lo que escriben tus agentes. Ninguno te avisa cuando dos de ellos creen cosas opuestas. Memanto es el primer agente de memoria: → Extrae decisiones y preferencias → Fusiona duplicados → Resuelve contradicciones → Retira lo obsoleto → Prepara a cada agente Open source, licencia MIT, un pip install. Tenéis el repo abajo :) @moorcheh_ai #adfrozen16d ago126K366272901250.43%
ALGUIEN HA CREADO UNA HERRAMIENTA QUE DOBLA CUALQUIER VÍDEO DE YOUTUBE A 100+ IDIOMAS CON LA VOZ ORIGINAL Se llama Voice-Pro y hace esto: → Descarga el vídeo → Elimina la voz limpiamente → Transcribe todo → Traduce a 100+ idiomas → Clona la voz original → Redobla el vídeo entero En menos de 2 minutos. 100% local y gratis. Os dejo el repo en los comentarios :)frozen5d ago124K3.1K62972233.6K5.97%
ANTHROPIC LANZÓ EL EMPLEADO MÁS BARATO DEL MUNDO Y CASI NADIE SE ENTERÓ Se llama "Claude for Small Business". Lo que puede hacer: → Gestionar facturas y finanzas → Crear campañas y contenido → Organizar ventas y clientes → Gestionar emails y calendarios → Ejecutar tareas entre apps Cómo funciona: → Conectas tus herramientas → Entiende todo tu negocio → Ejecuta flujos automáticamente Funciona con Microsoft 365, Google Workspace, Canva, QuickBooks y más. La idea es simple: en vez de abrir 10 herramientas, hablas con Claude y él hace el trabajo. Dedícale unas horas, lo agradecerás.frozen20d ago105K648954438311.54%
I've spent way too long trying to describe a camera move in a prompt, only to get something completely different. Turns out, you can bring motion from a reference video into your own creation with motion capture in Dreamina Canvas. A full month of Dreamina Basic is $1.50, with credits for up to 9 ten-second Seedance videos or 393 GPT Image 2.5 images. Dreamina 2.0 is canvas-first, the new way AI video is made. 👉https://t.co/ATDFQgbCQy #Dreamina #DreaminaPartnerfrozen3d ago101K19019270330.27%
my agents already live in my inbox and my calendar. they still have no idea what's on my desk. a personal agent with eyes flips that. per the thread, the hard part is "this": knowing which thing you mean when you point at it. that's the bar NewEyes AI has to clear."frozen10d ago92K15515130270.23%
Dreamina has never dropped a deal this big: a full month of Basic for just $0.29. Not a trial. Not a typo. I'm proud to be one of 1,000+ creators officially chosen for it and the $0.29 price only unlocks through an official creator link, so use mine below. I know what you're thinking. There's gotta be a catch, right? https://t.co/phEC9PT1Ta #DreaminaOnly29Cents #Dreamina1d ago91K21518100240.29%
UN DESARROLLADOR CHINO CONSTRUYÓ UNA HERRAMIENTA QUE HACE EL TRABAJO DE CUALQUIER EDITOR. Se llama MoneyPrinterTurbo y ya tiene +83.000 estrellas en GitHub. Escribes un tema y te devuelve el vídeo terminado: guión, clips HD, subtítulos, música y voz en off. El mismo pipeline que usan los canales faceless que facturan 6.000-10.000$/mes. Ellos pagaron. Esto es gratis. → Funciona con Claude y GPT → También DeepSeek y Gemini → Varios vídeos a la vez → 9:16 para TikTok y Reels → 16:9 para YouTube → Voz con preview en directo → Subtítulos 100% editables → Música ya integrada Corre en local, o sin instalar nada con Google Colab. Sin suscripciones. Sin marcas de agua. Sin límites. Enlace abajo :)frozen24d ago89K8181224141.1K2.38%
GAMMA JUST CHANGED HOW PRESENTATIONS GET MADE @GammaApp launched Gamma 5 today, and they rebuilt how it thinks, designs, and edits. → The agent asks first → Imports your PowerPoints → Exports with formatting intact → Pulls from Notion and Slack → Design styles that fit My biggest problem with decks was never the content. It was fighting the layout. I've been testing the Gamma 5 beta, and this is the first time it actually understood what I was building before building it. Old decks stay saved, and Gamma Classic is one click away. Here's a real deck I made with it 👇🏼frozen16d ago87K31115170310.43%
UGC CREATORS JUST GOT THE WEIRDEST COMPETITION IN THE WORLD A duck. An elf. An alien. And they're already shooting ads. A UGC creator: → Needs the product shipped → One face, one niche → Every reshoot is a new invoice An AI character: → Needs a prompt → Any face, any niche → Every reshoot is a new generation Brands don't really care who's in the ad. They care that it stops the scroll, and that there's a fresh one ready next week. 🔖 Save this.frozen17d ago85K26726340590.45%
STOP COPY-PASTING BETWEEN YOUR AIs. NOW THEY TALK TO EACH OTHER. Until today you were the messenger. Copy from Grok, paste into Claude, copy from Claude, repeat. @BolterApp just fixed that. Here's how they connect: 1. Copy one line 2. Paste it into your AI 3. Confirm with a code 4. It's in the chat Then you share its handle and it starts working with your other AIs and your friends' AIs. Give them a task and they split it up. One picks the landing site, another designs the wheels, another writes the code. The rover is ready and you didn't copy a thing. Each AI keeps its memory and you approve every connection. Link below 👇🏼frozen9d ago85K26121341870.48%
SAME VIDEO. ONE OF THEM COST 9X LESS. Top is WAN 3.0. Bottom is Seedance 2.5. Honestly, I couldn't tell which one was which until I saw the labels. The difference is the price: → WAN 3.0 is up to 9x cheaper → Now 60% off on the API → And unlimited inside Higgsfield That last part changes everything. The expensive part of AI video was never the final clip. It's the 30 failed tries before it. Now those are free. Link below ⬇️frozen9d ago67K20429250560.47%
🚨 ¿Y si el siguiente motor de videojuegos no se programa, sino que se genera? PixVerse R2 crea el mundo en tiempo real mientras tú te mueves dentro. → Te desplazas con WASD → Controlas la cámara → Escribes un prompt y el mundo cambia → Sin esperar a que se renderice nada Todavía es pronto, pero la dirección está clarísima.frozen29d ago63K491303301291.09%
@naturewill trained its first general model for the physical world, Natura-1, with 1 million hours of egocentric data. Regular people doing regular things with their hands. And now the robot: → Puts glasses in their case → Folds a towel → Holds something in one hand → While grabbing another object That's not even the crazy part. It does tasks nobody taught it. Ask it something new and it adapts on its own, no retraining needed. Grabbing an object is easy. The hard part is everything your fingers do without you noticing. That's what Natura-1 has learned. Watch the video 👇🏼 https://t.co/0AqRR7O7WQfrozen10d ago58K17418200320.42%
CHATGPT'S INTELLIGENT UI IS COOL. BUT IT ONLY WORKS INSIDE CHATGPT. Building your own AI app? Until now: → Your agent does the research → Then dumps a wall of text OpenUI 1.0 changes that: → Charts, forms, dashboards → Inside your own product → Using your own components → Works with any model → Fully open source ChatGPT has it. Now every app can 👇🏼frozen7d ago55K14418151930.49%
I DROPPED A CAMPAIGN BRIEF INTO AMBI. IT STARTED THE WORK ON ITS OWN AND ONLY CAME BACK FOR MY DECISIONS 👇🏼 A brief usually means reading it end to end, highlighting it and copying what's pending into another app… and something always slips through. This time Ambi read it and got moving: → Pulled the key points → Found the next steps → Started on them itself → Asked me only for decisions If I discuss the brief in a meeting, I can actively turn on Listening to capture the key points and get a recap afterwards. I can also set a Routine to keep watch for the deadlines or unfinished tasks I specify, so I don't have to repeat the same instruction every day. It works within the permissions I grant, from the information I choose to share or connect, and asks for confirmation before taking consequential external actions. It's available today on iPhone, Mac and Apple Watch, with its first hardware planned for early 2027. You can download Ambi and try it for free, link below :) @hello_ambi #ambi #AmbiLaunch #adfrozen10d ago52K13915240320.40%
LE PASÉ MI PLAN DE CONTENIDO DE LA SEMANA A UN EQUIPO DE AGENTES EN TEAMILY AI. Y A MI EQUIPO, EN EL MISMO GRUPO. Mi problema de siempre: un chat investiga, otro escribe, y yo copiando contexto entre los dos y reenviándoselo todo a mi equipo. En Teamily AI esto funciona como un equipo Humans + Agents: personas y agentes en el mismo grupo, con el mismo contexto. → Market Research busca los temas → Content Creator parte de eso → Saca los hooks → Todo en un doc compartido A mitad escribí en el doc: "esta semana, solo vídeo con IA". Rehicieron el plan entero sobre lo que ya tenían, y acabó en una web hecha con Website Builder. Y cualquiera del equipo puede entrar al grupo, pedir un cambio y los agentes lo aplican sobre la misma web. Sin reenviar nada. La diferencia con un chatbot: personas y agentes trabajan sobre el mismo contexto, de principio a fin. El cambio entra una vez y se aplica en todo. @Teamily_AIfrozen14d ago52K34824360560.89%
🚨THIS JUST KILLED MUSE, GROK BOT AND INSTINCT IN ONE GO It's called Fo MCP and it plugs straight into the AI agent you already use. It allows you to: → Make real phone calls from your terminal → Book flights, tables and appointments → Find a time and send the calendar invite → Shop a full look from one photo → Control your lamp, smart bulbs and robot vacuum No new app. No switching tabs. Your agent plans it. Fo does it. Every AI agent in the world can now act in the real world. This is how one person builds an entire business. 🔖 Save it.frozen3d ago52K243252301430.84%
i just read the sign out loud and had to put my phone down for a second https://t.co/lSvgZG5fKXfrozen15d ago51K24730230390.66%
NUEVO MODELO DE NEGOCIO: LOVABLE + BREKEN Hoy cualquiera monta una app con IA en un fin de semana. El problema llega cuando entran los primeros clientes y algo se rompe. Ahora la construyes con Lovable y antes de lanzarla la pasas por Breken: → Simula miles de usuarios → Mete años de datos → Hace fallar los pagos → Tira los servidores Encuentra lo que se rompe, lo arregla y demuestra que funciona. Construyes rápido y lanzas tranquilo.1d ago51K246292543861.35%
The hardest part of a literature review was never the reading. It was comparing four papers without losing track of which one said what. Dropped four open-access PDFs into ScienceBuddy and asked what evidence they gave on immunotherapy resistance. → No code, no local setup → Plain language question → Every claim traced back to its paper → Table downloads straight out Free, from @PhAILabs, at https://t.co/fVjww2lSD3 Biomedicine only for now, and sign-up needs a .edu address at the moment. ScienceBuddy is currently a Preview release. Verify results yourself before using them in research. @PhAILabs #ScienceBuddy #PhAILabs #AIforSciencefrozen29d ago51K31326300340.80%
ESTA IA CONVIERTE TUS PDFS, EXCEL Y ESCANEOS EN DATOS ORDENADOS A la izquierda, tus archivos: contratos, manuales, facturas. A la derecha, cómo los ve la máquina. Es Nace Document Intelligence, lo nuevo de @NaceAI: → Lee 36 tipos de archivo → Ordena y extrae los datos → Cada dato con su origen → A una fracción del coste → Con CLI y SDK incluidos Y se integra fácil con Claude Code, Codex, PI, Opencode, Hermes y Prime Agent. Antes de que una IA pueda decidir algo, primero tiene que entender tus documentos. Esto hace justo eso. Según su propio ranking, empatado en el #1 en lectura de documentos. Link abajo :)frozen14d ago50K19927240470.59%
thing I'm most curious about with ElevenAgents Architect: whether it holds up on a big messy multi-step workflow or only on tidy ones. ours is not tidy. trying it tomorrowfrozen15d ago49K2721880280.67%
2024: hire a marketing team 2026: go to bed and let Selfdrive get you users https://t.co/xRA2D0aSxQfrozen9d ago49K11814140360.37%
YOU CAN NOW DIRECT AN AI VOICE LIKE AN ACTOR. Not "happy" or "sad" presets. Real directions, in plain words: → "Barely holding it together" → Whisper just one word → Sigh, laugh, breathe → Calm to panic mid-sentence Drama 3 from Fish Audio. And the weights are open. Voice AI just got a director's chair.frozen9d ago48K11117131370.37%
YOU USED TO NEED DAVINCI AND YEARS OF PRACTICE Now you just describe the look to your AI agent. @trymirage just added professional color grading to Tesseract: → Grade log footage → LUTs, ACES and HDR → Color wheels and curves → Masks for precise fixes → Match every single shot From subtle correction to a full cinematic grade, without leaving your project. And every adjustment stays editable, so you can keep refining it yourself. This is moving soooo fast.frozen8d ago48K13316190360.42%
🚨 ¿ENTIENDES LO QUE ACABAN DE LANZAR? Un modelo que no escribe. Solo decide. Qué herramienta usar. Si un resultado vale. Cuándo pasarle el trabajo a un modelo grande. Y según sus propias pruebas: → Gana en 13 de 16 → 96,3 en Decision RULER → Con 1M de tokens → 74 vs 48 en LCR Los primeros 10.000 registros tienen 100M tokens gratis. Esto se va a llenar rápido.frozen10d ago48K21620170840.71%
Same shopping agent. 20 decisions in a row. One already bought the item → Drex: 20/20 done → Microsoft Decision-1: 18/20 → OpenAI GPT-6 Sol: 3/20 Drex placed the order in 10.4s. GPT-6 Sol was still stuck on decision 4. Drex is the fastest decision model on @OpenRouter right now. By @NaceAI 👇🏼frozen12d ago47K966220270.32%
LOS AGENTES YA ESCRIBEN TU APP ENTERA. EL PROBLEMA ES PONERLA ONLINE. InstaCloud es la nube pensada para agentes: → Tu agente despliega directo → Escala solo con tráfico → Baja a cero sin uso → Copias con datos incluidos → Cero DevOps manual Pagas solo lo que consumes, y la capa gratuita incluye $10 al mes de uso.frozen23d ago47K17121140430.53%
YOUR FRIEND TOLD YOU TO BUY NVDA. HERE'S HOW TO CHECK IF HE'S RIGHT BEFORE YOU PUT IN A SINGLE DOLLAR We've all been there. Someone you trust says "just buy it", and you either go all in on vibes or spend hours reading earnings reports you don't fully understand. So I tried something different. I asked Minara, an AI built only for investing, to do the homework a Wall Street analyst would do: → Pulled the latest earnings and filings, with the date of every source → Separated real facts from management guidance and assumptions → Gave me the bull case AND the bear case → Listed 3 things that would prove each side wrong That last part is what changed how I think about it. Instead of asking "should I buy?", you start asking "what would make me change my mind?" Then I opened its sources to check the numbers myself. Don't blindly trust your friend, and don't blindly trust an AI either. Next, I opened Trading Co-pilot to review the entry, position size, and exit levels. On NVDA spot with $100: entry $229.54, target $234.50, invalidation $226.45. Its call? Hold. Good company, bad entry today. Full video below :) #MinaraAI #FinancialAI #AgenticTrading #AgenticFinancefrozen29d ago47K33022170310.86%
NO OTHER LIVE AVATAR I'VE SEEN COMES CLOSE This is Jesus hosting a late-night podcast, running live on Vivix A1. You can talk over him mid-sentence and he adapts. → Full-body gestures → Real interruptions → Scene stays coherent → Zero turn-taking Most avatars are a talking head on a video call. This one's actually in the room. @VivixLabs_HQfrozen22d ago47K17627090.41%
PREDICTION MARKETS ARE FUN NOW The newest prediction market is more like a game. It lets users rip open packs to get access to markets and constantly see what’s popular. You can trade in business/tech and even pop culture. It’s designed for people who love trading and for those who have never used a prediction market before. Everyone should give it a try right now. https://t.co/Rd9u3nUTkx1d ago47K7711150270.28%
ATTACKERS USED TO NEED TO BE CLEVER. NOW THEY JUST NEED TO BE TIRELESS. A human gives up after a few dead ends. An agent swarm doesn't. Cogent's research found that attack paths only AI agents would bother with run at least twice as deep as the ones humans go for. So defense has to change: → Map every path → Verify every hop → Find the chokepoint → Cut it before they arrive14d ago47K7415100260.27%
I just tried Dreamina Seedance 2.5 (for preview), and it honestly changed how I approach AI video testing. Instead of spending HD credits every time I want to try a different camera move or action, I can draft first at 480p, find the version I actually like, then generate the final 1080p from that exact take. The best part? It's not an upscale. The final video is generated natively. Also found out Dreamina Basic is currently $1.5/month, which is kind of wild. More on that below. #Dreamina #DreaminaPartnerfrozen26d ago47K30618130300.79%
I manage 10+ X accounts and 60+ creators. Every generic AI tool I tried made me adapt my workflow to it. Never the other way around. Workmate flips that: custom agents built around how your company already works, plus engineers who maintain them. Tool vs. teammate. https://t.co/z8DedvOH3xfrozen8d ago46K991890240.32%
Dreamina AI updated its web experience and canvas, and it’s live now. @Dreamina_ai #DreaminaCaughtTheVibe The update brings high quality AI video workflows and skills for different creative needs, including film, brand ads, viral social media content, and more. It makes it easier to explore ideas and move from concept to content in one place. Get 90% OFF your first month of Dreamina Basic until October 9th. Try it out on your next creative project.frozen28d ago46K18410140220.50%
OPUS 5.5 + BOREAL-H3 ES UNA LOCURA Una sola petición. Un solo intento. Y salió este anuncio de Olipop. → Misma lata → Misma etiqueta → Mismos colores → En cada plano La mayoría de modelos de vídeo hacen vídeo creíble. Este hace vídeo correcto. Pruébalo en Model Playground ↓frozen22d ago46K17013110320.49%
AI video usually looks impressive until you ask it to actually move. Fast camera movement. Multiple subjects. Large-scale action. That's where things usually start falling apart. Kling 4.0's Dynamic Motions is built for exactly this kind of shot. https://t.co/s35V4gWYeDfrozen23d ago46K32723150330.87%
I TRIED AMBIENT BY @vellum_ai TO BUILD A CAMPAIGN REPORT IN GOOGLE SHEETS Double-tapped Fn, shared my screen and asked "help me get the average CPM per account from this sheet" What happened: → Spotted views stored as text → Showed me on screen where to fix it → Dictated the formula, I pasted it with hold Fn About 4 minutes to a finished table. By hand it usually takes me 15 Not perfect: once it annotated the wrong column and I had to correct it by voice. It got it right away For screen work where you'd normally take screenshots and paste them into a chat, this saves a lot of timefrozen28d ago46K1515100130.39%
YOUR AI AGENT CAN NOW UPGRADE ITSELF. Reef just launched /reefine: you describe what you want and it builds the change, tests it and ships a new version on its own. → Adds new modes → Tells you out loud → Integrates new tools → Works with Codex, OpenCode... Basically: your agent evolves just by talking to it. It's open source. Link below :)frozen27d ago46K12024121440.44%
my roomba could neverfrozen23d ago46K1261380190.36%
I HIRED A HEAD OF CONTENT WHO NEVER SLEEPS, NEVER COMPLAINS AND NEVER TAKES A DAY OFF Every morning at 8:00, his message is already waiting. I used to start my day opening X with no plan. Now Stanley sends me every morning: → Trends in my niche → 3 post ideas → What competitors did → Replies to answer → People to DM I just reply: → "Yes" → "Change the hook" → "Schedule it for 6pm" What used to take me 2 hours now takes 5 minutes. Here's what the chat looks like 👇🏼 Here is the link: https://t.co/Xi8p7xhFPMfrozen22d ago46K33721160390.91%
NADIE HABLA DEL VERDADERO PROBLEMA DE LOS VÍDEOS CON IA: EL AUDIO. Generas un clip perfecto, pero el sonido no pega con nada. Y hasta ahora solo había una salida: regenerar el vídeo entero. Mirelo SFX 2.0 cambia eso: → Mantienes tu vídeo → Cambias solo el sonido → Quitas la música sobrante → El diálogo no se toca Ya no hace falta tirar un buen vídeo por un mal sonido.frozen8d ago46K12215180480.45%
Holy shit... the gap between what AI video tools promise and what they actually hand back just got a lot smaller. Normally, using AI for video feels like a gamble: you type in a prompt, wait three minutes, and it spits back a locked MP4. If one single specs number on screen is wrong, you can't edit it—you just have to re-prompt, re-render, and burn through your credits again. To see if any tool could actually handle a real commercial asset without the usual headache, I ran a quick test. I took a concept photo of an "iPhone Duo" and asked @fotor_com’s Fotor Agent to build an Apple-style product reveal—no 3D modeling, no physical camera setup, and no hiring a motion designer. It didn't just render a clip. It handed me an actual multi-track project file in about 40 minutes. A few things genuinely surprised me about the workflow: 1. It built a real timeline (and did its own research) I uploaded the concept shot and asked for a reveal with animated specs. I didn't even feed it detailed technical specs—it actually went online, pulled plausible data, and mapped it out. I cross-checked against Apple’s actual site later and the formatting and spec logic were spot-on. Instead of a flat video, it laid out the clips, music, transitions, and motion graphics onto separate editable tracks. 2. Native 4K Motion Graphics for a few cents a second Getting clean, keyframed motion graphics usually means opening After Effects or paying someone a few thousand bucks to build them. Fotor Agent generated sharp, native 4K spec cards and charts right on the timeline. The cost broke down to just a few cents per second, which makes testing complex visual ideas almost risk-free. 3. You can fix mistakes without re-rendering everything This was the most practical part. When I wanted to shorten a piece of copy on one of the spec cards, I didn't have to generate a new video or burn more credits. I just clicked the text layer directly on the timeline, edited it, and the preview updated instantly. The actual breakdown: Time: 3–5 days of AE keyframing → ~40 minutes Cost: Thousands in agency quotes → A few dollars in credits Workflow: Jumping between 5 different apps → Done in one Agent AI shouldn't just dump a locked video file on your desktop and call it a day. Fotor Agent handles the heavy production setup while leaving you with an actual project file you can tweak and refine. Screen recording and the 4K test clip are attached below! 👇🏼frozen23d ago46K26624230440.78%
Your AI agent handles thousands of requests every day. And it learns absolutely nothing from them. Reef wants to change that: open-source infra that lets agents keep improving from real usage 🧵 https://t.co/4TRnvKTXbafrozen24d ago46K20725230400.65%
AN AI THAT FITS ON YOUR LAPTOP JUST WENT UP AGAINST MODELS 33X ITS SIZE Here's how it did: → Only 2 GB → Works offline → No usage fees On math problems, it lands just 3 points behind a 120-billion-parameter model. It has 3.6 billion. On pure logic, it ties with a model twice its size. It's called TwIL-LM3-Pro, built by @thewebai. Link below 👇🏼 @Davidstoutfrozen22d ago45K1774150240.49%
MINARA JUST LAUNCHED AN API WITH OVER 50 MODELS IN ONE PLACE Minara Cloud gives you access to GPT, Claude, DeepSeek, Kimi and 50+ more: → One single API → One single bill → Plugs into your agent → 20% bonus on every top-up It connects to Hermes Agent, omp, OpenCode or anything that speaks the API. No need to change your stack. Try it now.frozen21d ago45K3048190210.78%
THIS AI SORTS YOUR SUPPORT TICKETS IN 1 SECOND In their demo, Drex reads each customer message and decides what to do: solve it on its own, pass it to a team, or alert a person. One customer asked to delete all their data and it sent it straight to a senior agent. → Reads the customer's message → Decides in one second → Only bothers humans when needed It's now open source and the first 10,000 people to try it get 250M free tokens. Link below :)frozen16d ago45K28220221600.86%
IF YOUR AI APP USES THE MOST EXPENSIVE MODEL FOR EVERYTHING, READ THIS. RouterPlus just launched and does the work nobody wants to do: → Tests models on your usage → Picks the most efficient one → Finds the best provider → Negotiates the price for you → Keeps a backup ready Real case: same model, same traffic, and the provider offers came back about 2x apart. They're offering a free review of your AI spend 👇🏼frozen9d ago45K851990340.33%
The product is fictional, but the comments asking where to buy it should be very real. #TheWorldsCatchiestAd #ElevenCreativeCompetitionVol1frozen17d ago44K12715100200.39%
I HIRED A HEAD OF CONTENT WHO NEVER SLEEPS, NEVER COMPLAINS AND NEVER TAKES A DAY OFF Every morning at 8:00, his message is already waiting. I used to start my day opening X with no plan. Now Stanley sends me every morning: → Trends in my niche → 3 post ideas → What competitors did → Replies to answer → People to DM I just reply: → "Yes" → "Change the hook" → "Schedule it for 6pm" What used to take me 2 hours now takes 5 minutes. Here's what the chat looks like 👇🏼 Here you've the link: https://t.co/Xi8p7xhFPMfrozen20d ago43K31117170370.88%
Okay, I had to double check this because $1.5 for AI video usually means you're talking about a trial, a tiny credit pack, or some weird first-time offer. But Dreamina is actually doing a full month of Dreamina Basic for $1.5. It's not a 3-day trial. It's the monthly plan, with the Seedance lineup included, including Seedance 2.5. Sep 23–Oct 9, up to 90% off. #Dreamina #DreaminaPartnerfrozen28d ago43K23620100340.69%
blind-test benchmarks against Meshy, Tripo, Cartwheel and DeepMotion. from Viggle. I still had them filed as the meme app, so reading the JST-2 research posts was a bit of a surprise. there's a proper technical write-up and they already name JST-3. https://t.co/shrK114Hoyfrozen3d ago43K1321790680.52%
LOS FOUNDERS PIERDEN SEMANAS ESCRIBIENDO LA MISMA HISTORIA 40 VECES. YC. Techstars. a16z Speedrun. Antler. Cada formulario pregunta lo mismo con otras palabras. Esta app gratis para Mac lo hace al revés: → Lee tu deck y tu web → Te entrevista una vez → Rellena 40+ aplicaciones → Solo para si no sabe Nada se envía hasta que tú pulsas Submit. Construye. No reescribas. https://t.co/Cv4vIz8NZ9frozen9d ago43K10619120440.42%
HAN JUNTADO UNA IA DE CIBERSEGURIDAD CON UNA HERRAMIENTA QUE TE ENSEÑA CÓMO FUNCIONA POR DENTRO CUALQUIER APP El combo es OrcaCyber Zero 1.5 + REA. REA abre la app y te muestra lo que hay dentro, aunque no tenga el código público. OrcaCyber Zero 1.5, una IA entrenada solo para ciberseguridad, revisa todo eso y busca fallos antes que los atacantes. Sirve para: → Programas de ordenador → Apps de móvil → Páginas web → Aparatos electrónicos Lo que hace OrcaCyber: → Encuentra fallos ocultos → Comprueba si son reales → Prioriza los más graves → Lee proyectos enormes No está abierta a cualquiera: hay que verificar tu identidad para usarla. Link abajo 👇🏼frozen10d ago43K353483612241.56%
ESTA IA USA TU ORDENADOR COMO SI FUERA TÚ. Se llama Incredible y no se queda en responder, hace la tarea entera: → Hace clic por ti → Escribe en tus apps → Navega por webs → Rellena formularios → Mueve archivos enteros Algunas cosas que le puedes pedir: → Meter facturas al ERP → Limpiar duplicados del CRM → Etiquetar tickets de soporte → Reprogramar tus reuniones → Migrar docs de Notion a Drive Le das una frase, te vas a comer y cuando vuelves está hecho. Funciona en Mac y Windows. Link abajo :)frozen16d ago42K536612603422.28%
SCIENCE IS NO LONGER JUST FOR PEOPLE WITH A LAB BEHIND THEM. AIPOCH Open-Science is an open source AI workbench where you give it a question and it runs the whole research loop without you switching apps. → Reads papers and runs the code → Connects to genomics and chemistry → You pick the model → Local by default → Every result is auditable The era of "Vibe Research" is arriving. Anyone with curiosity or a research idea can now take part and connect with researchers around the world through open source tools. The good part isn't that it's fast. It's that you can see where every number came from. And now all of it fits in one file: .science → The full conversation → The data you used → The code that ran → The final charts Send it and anyone can see how you got there. Best example I've seen: Euler claimed in 1769 this had no solution. It took until 1966 to prove him wrong. They recreated that search in Open-Science: → 1 million tests: nothing → 1 million random tries: nothing → One smart search: 0.082 seconds 27⁵ + 84⁵ + 110⁵ + 133⁵ = 144⁵ The whole process is in one .science file you can open yourself. ⭐ 4.6k stars 📄 Apache 2.0 🔖 Save it. @aipoch_ai #AI4Science #OpenSourcefrozen8d ago42K8414172350.36%
English and Spanish, same voice, same emotion on Eleven v4 in this clip. The feeling survives the language switch, which is the part that never worked before. https://t.co/Fw8xznJgBsfrozen18d ago42K17520131330.58%
STOP DOWNLOADING APPS FOR EVERY LITTLE THING. A trip. A movie night. Marathon training. Splitting bills. Each one used to be a different app. → One chat → Your friends inside → AI builds what you need → Gone when you're done Wild how fast this is moving. https://t.co/E0t1zLxGt6frozen3d ago42K1161570730.51%
DISNEY IS COOKED someone in their bedroom can make this now what the fuck are we doingfrozen26d ago41K13011160330.46%
AI ANIME CREATION IS ENTERING A NEW ERA. 🎬 With Vidu Q4 Preview, experimenting with anime-style animation doesn't have to break the bank. At just $0.014 per second, half the price of other models, you can bring more ideas to life: → 💥 Explosions and particle effects that bring battles to life. → 🎥 Dynamic camera movements that follow the action. → 🎨 Anime-inspired visuals with 2K and 4K output. → 🖼️ Up to 15 reference images to guide your creations. → ✨ More consistent characters across different shots. The real breakthrough isn't just lower costs. It's the freedom to experiment, scrap ideas, and try again until your vision comes together. From AMVs and fan animations to original anime-style stories, there's more room than ever to explore your creativity. 👇 Check out the example below! 💬 If you could recreate any anime moment with AI, which scene would you choose, and what would you change about it? #ViduQ4Preview #ViduAI #Anime #AMV #AIAnimation #AIVideo #AnimeCommunity #AnimeCreator #adfrozen6d ago41K84115120342.20%
this is exciting how ai is now getting into film and production founder's time is very precious and he/she should spend it wisely in tho things that can add revenue. and in this journey when you give 5 take for a 20 sec reels you're loosing not just time but money as well @trydrama_io will be a helping hand ig for them.. this seems something exciting happening in ai filming industry.frozen10d ago41K13722170430.54%
AN EX-APPLE PM JUST LAUNCHED HER OWN HEALTH AGENT Praneeta Pujari: → 6 years on Apple Watch → Lives with migraines → Cares for her mum She's spent over 10 years managing her mum's chronic pain, and that's where Niq comes from. An agent that connects your watch, your records and your labs to look for signs of a migraine flare before it hits. And it doesn't stop at alerts: → Books the appointment → Sorts the prescription → Handles the insurance → Delivers your meds With real doctors behind it.2d ago40K7917160290.35%
🚨 ÚLTIMA HORA: EN GTA YA PUEDES DECIRLE A UN NPC QUE SE SUBA A TU COCHE Y tener una conversación mientras conduces. No es un guion. Nadie ha escrito sus respuestas. → Tiene voz propia → Tiene personalidad → Se acuerda de ti → Sabe lo que pasa en la ciudad Más de 300.000 turnos de conversación en tres semanas. Esto ya no es un videojuego. Es una ciudad.frozen25d ago39K284443721081.20%
NOBODY OWNS THE WHOLE OPEN AI STACK. So DigitalOcean put it in one room with NVIDIA. The Open Intelligence Summit today: → Hugging Face on what's next → vLLM serving for agents → Open harnesses with Nous → Shipping on an open stack Bolt, Supabase, OpenRouter, MiniMax and more on stage. Live from 1 PM PT.frozen9d ago39K8015120280.35%
Which platform did you join because "you have to be there"? Still there?frozen7d ago39K1692080250.57%
SIGAN A ESTA CUENTA, NO DIRÉ NADA MÁS…3d ago39K20530270610.84%
LA IA PUEDE RESPONDER MUCHO MÁS RÁPIDO GRACIAS A ESTO Se llama EAGLE-4 y acaba de salir en GitHub como source-available. Hace que un modelo de IA escriba sus respuestas más rápido usando el mismo ordenador. Sin comprar hardware nuevo. Según los resultados de su paper (Qwen 3.8 27B, 4× B300): → 3,59× más rápido → Mismo equipo → Mismo modelo → Supera a sus rivales Y aguanta incluso con 512 personas usándolo a la vez. Os dejo el repo abajo 👇🏼1d ago39K194142211230.92%
SIGAN A ESTA CUENTA, NO DIRÉ NADA MÁS…2d ago38K1021370530.46%
CLAUDE + HIGGSFIELD UGC ADS IN 8 MINUTES Andy Lo shows the whole system: Build an AI character. Let Claude write the video prompts. Batch them all at once. The article below covers what comes next: scheduling it all with Postiz. 🔖 Save it. https://t.co/5TsiB0bkNj15d ago38K32641413771.27%
finance bro bragging about his agent memory 😂frozen7d ago38K591590200.27%
WHAT IF EVERYONE HAD A PERSONAL DOCTOR? That's Nolla Health's ambition, and they're starting with something very specific: acne in Utah. → AI-guided consultation → Diagnosis and treatment → Prescription when appropriate → Clinical follow-up after Today it's one condition in one state. The goal is to reach everyone.frozen17d ago38K13221130400.54%
Esta semana vi tres pitch decks hechos con IA y los tres tenían el mismo degradado morado. La gente lo nota en la primera slide. El problema no es usar IA. Es conformarte con lo primero que te da. Os presento Gamma 5 👇🏼 https://t.co/IrIqY6IdV1frozen10d ago38K14920190450.62%
our investor update video takes a full week every quarter just to look like the last one. curious what Syren does with thatfrozen13d ago38K6713100270.31%
ENTERPRISE SOFTWARE IS ABOUT TO LOSE ITS SCREENS https://t.co/usEyv5aJPn just launched something that changes who uses software: → Describe your process → Proto builds the app → Proto works inside it → You just approve CRM, payroll, warehouse, finance or internal workflows. All from a prompt. This is headless SaaS: the agent uses the software through APIs and you stop fighting with menus. Link in the first comment 👇🏼frozen7d ago38K19524190320.72%
GENERAL COMPUTE IS THE NEW INFERENCE CLOUD FOR CEREBRAS 20x faster AI, but almost no compute to run it. Watch the fastest inference. Track the compute shortage. Mark Q1 2027. Speed → Compute → Lower cost 🔖 Save this. https://t.co/ROLzhFm5Cffrozen23d ago37K2756160160.84%
ALGUIEN HA CREADO UN REPO EN GITHUB PARA QUE TODO TU EQUIPO USE TU CLAUDE CODE Hasta hoy tu agente solo hablaba contigo. Tu equipo quería un cambio y te lo pedía a ti. Tú se lo pasabas a Claude Code. Y vuelta a empezar. El cuello de botella no era la IA. Eras tú. Este repo le da a tu agente identidad propia. Un @, una bandeja de entrada y sitio en los canales del equipo. Lo etiquetas como a una persona más. → Tu socio le escribe por DM → Tu cliente le pide cambios → Recuerda lo que se habló → Sigue con lo pendiente Y el agente no se va a ninguna nube. Sigue corriendo en tu ordenador, sin suscripciones nuevas. Tu equipo le habla desde el móvil. Funciona con Claude Code, Codex, Cursor y OpenCode. Antes tenías una herramienta. Ahora tienes un compañero de equipo. Se llama Alook. 1.2k estrellas y open source.14d ago37K14624440850.81%
this one actually scared me. someone clones a rich client's voice and calls her advisor from her own number. the advisor can't tell. nobody can. this is happening right now. built this with DetectifAI's API this morning. client of 10 years asks to wire $400K. voice check says it's her. caught as fake in 6 seconds. money stays put. the fake voice is the founder's own. they've solved it. API open to everyone today. @DetectifAI3d ago37K10023160420.49%
ODYSSEY JUST LAUNCHED AGORA-2 AND IT'S A GAME ENGINE LEARNED BY AI → 20 humans and agents in one live world → 5x more players than Agora-1 → Multiple environments, not just one → Every action changes the game for all The bigger picture: this is how robots, self-driving cars and cyber agents will learn to cooperate and compete, safely in simulation. You can try it todayfrozen28d ago36K2445151150.77%
AI JUST TURNED TASTE INTO A PAYCHECK Here's how Demoverse works: → Pick a product → Improve it with words → Best design wins → You earn royalties Brands validate before manufacturing. You get paid to decide what gets made. https://t.co/YzFl7XYWB5frozen16d ago36K2781490270.91%
A 5-SECOND AI VIDEO IN UNDER 2 SECONDS. From $0.02 per second. @PrunaAI's new P-Video-2-Pro Reference: → Lock a product or character → Feed it images, clips, voices → Get back an MP4 with sound → 480p or 768p Built on MiniMax H3 Ref2V. Every frame of this clip came from it.frozen7d ago36K571280250.28%
CRM cleanup is the task I've wanted to hand off forever. my Cursor setup could write the code but couldn't log into anything. going to try it through SpringBrand this week. Exactly the kind of chore I'd love to never touch againfrozen24d ago35K1912060290.69%
JOB HUNTING IS NOW A SINGLE PROMPT. “find me a better job, remote, AI engineer, get the top 10 ready, apply for me” no browsing job sites and no endless tabs @dreamworkhq finds the roles and writes each application. you say send, and an agent submits the application directly on company sites14d ago35K14517120320.58%
THIS STARTUP JUST MADE GPT-6.1 FREE No subscription. No API key. No card. Freebuff runs coding agents paid for by ads. → GPT-6.1 Sol daily → GPT-6 Luna → DeepSeek V4.1 → MiMo 2.6 Pro One command: npm install -g freebuff 🔖 check here: https://t.co/CycsrlxjDh https://t.co/xLy9W8CRRWfrozen19d ago35K13416210770.70%
the interesting number here isn't the 18%, it's the 700x. an eval you can afford to run on every response is a different product from one you run on a sample. that's the part that changes how you build.frozen27d ago35K1209120180.46%
extremely important if you run a fintech company:frozen28d ago34K2718181150.91%
me being the messenger between grok and claude since 2023 seeing this did it all on its own... use code AISC on @BolterApp and get 1 month free https://t.co/v7HYWX3zaTfrozen8d ago34K10416121240.46%
ACABAN DE LANZAR EL "SIGN IN WITH GOOGLE" PARA AGENTES DE IA Se llama AgentID y es de AgentMail. → Tu agente usa su email → Queda vinculado a ti → Sin compartir tu contraseña → Lo revocas y se acaba Para las apps es gratis y se instala con un comando. Si lo integras y respondes al hilo con captura: 3 meses de AgentMail gratis.frozen16d ago34K17913110480.74%
THIS IS GOING TO HURT A LOT OF AGENCIES A professional product video used to mean: → scriptwriter → voiceover artist → editor → motion designer Now it's one screen recording and 5 minutes It's called PuppyDog and 1,000+ companies already use it https://t.co/kuc0ZXCtWnfrozen29d ago34K19217131400.78%
Your agent stack already has evals. Your voice stack probably doesn't. There, the first eval is your listener. @OnepinAI checks each line for clarity and pronunciation and fixes mispronunciations directly. Here it is, script to voiceover: https://t.co/GPMOUeDKspfrozen21d ago34K1265100140.46%
$46.86 FOR 30 ELECTROLYTE STICKS. OR $9.35. That's Liquid I.V. on Amazon vs Aonic Race. Same 500 mg of sodium per stick. Here's where the rest goes, by Aonic's estimate: → Product itself: $5.91 → Brand takes a cut → Distributor takes a cut → Store takes a cut @AonicClub sells it at cost and only earns on the membership. @maxjmeier just opened it.frozen9d ago33K621560210.31%
AI CAN TEACH YOU LANGUAGES NOW. This model has the best pronunciation of all AI voice tech. It can teach you how to actually speak in a language the way native speakers do. As someone multilingual, this is insane. Drama 3 by @fishaudio is the most emotive model on the market right now. Try it with any language.frozen9d ago33K22423290671.05%
AI LETS ENGINEERING SHIP NEW FEATURES EVERY WEEK, BUT THE VIDEOS EXPLAINING THE PRODUCT ARE STILL MADE BY HAND. Clueso just raised a $10M Series A to fix that: → agents use your product → record the walkthrough → update it every release 1,000+ companies already on it. Link to test it for free below 👇🏼frozen9d ago33K12321130340.58%
OpenAI brain + Higgsfield creativity??? this combo shouldn't be legalfrozen23d ago32K7423250360.49%
LA MAYORÍA DE APPS HECHAS CON IA SE QUEDAN EN UN PROTOTIPO BONITO Myndlab te genera la app entera: → Frontend → Backend → Base de datos Y el código es 100% tuyo. Lo exportas, lo despliegas donde quieras y sigues construyendo encima. Incluso puedes elegir tu propio backend en vez de uno por defecto: Go, Rust, Python, .NET... Esa es la diferencia entre una demo y algo que puedes lanzar de verdad.frozen15d ago32K19321170780.96%
MAKET 2.0 TE DEJA REFORMAR TU CASA SIN SABER DIBUJAR PLANOS ¿Quieres otra habitación? ¿Una cocina más grande? ¿Otra distribución? Subes el plano de tu casa y se lo explicas a la IA con tus palabras. Te enseña cómo quedaría, y lo puedes cambiar las veces que quieras. → Sube tu plano actual → Pide otra habitación → Rehaz la cocina → Míralo en 3D Así llegas al arquitecto con algo que enseñar. 30 créditos gratis, sin tarjeta.frozen20d ago31K16119220410.79%
wait, why didn't this exist already???frozen17d ago31K1221090270.55%
BOLT IS GIVING $10,000 TO WHOEVER BUILDS THE BEST BUSINESS THIS WEEK And the best part: you can build it for free this weekend. → You have an idea → Turn it into a product → Submit it by the 20th → Compete for $10,000 Your project doesn't have to be new. If you've already built something on Bolt, it counts too. Link below 👇🏼frozen5d ago30K9312110230.47%
SILICON VALLEY'S FAVORITE $40 EMAIL APP JUST GOT AN OPEN-SOURCE RIVAL. Superhuman is the email client founders and VCs swear by. Fast, keyboard-first, and $40/month if you want the AI. Inbox Zero just launched its own email app built on the same idea: → Works with Gmail and Outlook → Free as an email client → AI for $20/month → Web, Mac, Windows and iPhone The AI is the interesting part: → Sorts every email for you → Drafts replies in your voice → Finds free slots for meetings → Set it up in plain English And it's open source. The code is on GitHub with 12k+ stars, so you can check exactly what it does with your inbox, or host it yourself for free. Not a weekend vibe-coded project either. They've been building it for 3 years. Why it matters: Your inbox is the most private thing you have. Being able to see the code behind the AI reading it is a big deal. Link below ⬇️frozen7d ago29K6711181280.43%
YOUTUBE PREMIUM NO TIENE NINGÚN SENTIDO DESPUÉS DE ESTO Hay una alternativa gratis que hace lo mismo y ni siquiera necesitas cuenta. Se llama Invidious, ya tiene 23.5K stars en GitHub y Google lleva dos años intentando cerrarlo. Te permite: → Ver sin anuncios → Audio en segundo plano → Ver sin que te rastreen → Importar tus suscripciones Ojo, las instancias públicas a veces fallan porque Google las bloquea. Si lo alojas tú va mucho mejor. 0€ al mes. 100% open source. 🔖 Guárdalo.frozen3d ago29K580912315644.31%
FORMING AN LLC USED TO MEAN 4 DIFFERENT WEBSITES. Now it's one flow. Form the company. Apply for the business account. Run everything from day one. → $349 all-in LLC, any state → State fee + registered agent included → EIN application filed for you → Invoices, bills, cards, expenses No more stitching tools together. 🔖 Save thisfrozen6d ago29K579110240.35%
IMAGINE YOUR COMPUTER KNOWING WHAT YOU’RE ABOUT TO TYPE Sabi is working on exactly that kind of future. → read brain signals without touching the scalp → predict the next 3–4 keystrokes → turn intent into text and eventually actions No brain implant required. $50M says investors think the way we interact with computers is about to change.frozen13d ago29K67990230.37%
okara acaba de meter a un experto en seo, un redactor, un community manager y un influencer marketer en una suscripción de $129 las agencias de marketing no están preparadas para esto https://t.co/jOR6kF6as8frozen3d ago29K1082080450.63%
El CEO de OpenAI cuando vio que un modelo con 84,5 en SWE-bench Verified se puede descargar gratis con licencia MIT y no puede hacer nada al respecto https://t.co/idNGZVsJP12d ago29K12315150540.72%
SI QUIERES QUE CHATGPT TE RECOMIENDE, MIRA ESTE ANUNCIO, NO DIRÉ NADA MÁS...frozen13d ago29K10319100600.67%
CONNECTING AI TO A COMPANY'S SOFTWARE IS HARDER THAN IT LOOKS. AND IT'S NOT JUST A TECH PROBLEM. AI agents need to work inside tools businesses already use, like Salesforce. When that connection is messy: → Engineers keep fixing it → Every new client takes months → Sales loses big deals Ampersand solves all three: → Less fixing → Faster client setup → More deals closed Already used by 11x, Orb, and Square.frozen16d ago29K13713120270.66%
🚨 ANTES DE COMPRAR UN SOFÁ, HAZ ESTO (O TE VAS A ARREPENTIR) Casi todo el mundo compra muebles a ojo y luego descubre que no caben en el salón. Maket acaba de sacar una función que lo soluciona: haces una captura de cualquier mueble y lo metes en el plano de tu casa con sus medidas exactas. Te dejo el paso a paso 👇🏼 1. Abre un proyecto: sube el plano de tu casa, genéralo con IA o empieza de cero. Luego pasa el ratón por "Objects" y pulsa "Create your own". 2. Sube una imagen: puede ser una captura de cualquier web o una foto que hayas hecho en la tienda. Ojo, solo funciona desde ordenador, no desde el móvil. 3. Selecciona el mueble: si la foto tiene varios, puedes elegir más de uno a la vez. Si solo hay uno, haz clic y listo. 4. Pon las medidas exactas: abre la ficha del producto en la web de la tienda y copia el alto, el ancho y el fondo en Maket. 5. Colócalo en tu plano: entra en "Objects", luego en "My Objects", selecciona tu mueble y arrástralo donde quieras. En 5 pasos sabes si ese sofá cabe de verdad antes de gastarte un euro. 🔖 Guárdalo, esto te puede ahorrar mucho dinero.frozen13d ago28K19835370981.30%
THE TOOL CALL RETURNED 200. IT WAS STILL WRONG. That's the agent bug regular monitoring never catches. The logs look clean. The user just leaves. DigitalOcean shipped a loop for exactly this today: → Insights traces every step → Signals flags the session → Simulations tests the fix → A person approves it All in public preview. Catch it at user one, not user four hundred.frozen10d ago28K9521110360.59%
SOMEONE FINALLY BUILT DATADOG FOR YOUR DOCS Falconer's Knowledge Health tracks: → Contradictions between sources → Coverage gaps → Redundant docs → Stale pages → Drift from code changes One score for how clean your company knowledge is. Agents are only as good as the context you give them.1d ago28K1318110190.61%
ESTA SKILL HACE QUE TU AGENTE DE IA TRABAJE POR MUCHO MENOS Si usas agentes de IA a diario, ya sabes dónde está el problema: cada tarea consume modelos, y la factura no para de crecer. AntSeed lo resuelve conectando tu agente a un mercado abierto de inferencia: → Una sola skill → Se configura solo → Muchos modelos disponibles → Precios bastante más bajos → Modelos gratis para empezar Sin cuenta. Sin API key. Compatible con Hermes, OpenClaw, Codex, OpenCode o tu propio agente. Cómo instalarlo, en el primer comentario ⬇️frozen9d ago27K18920231621.08%
121 PEOPLE. 36 REAL CALLS. NOT ONE VOICE MIXED UP. Luel Multi-Speaker Separation takes a call where everyone talks at once and gives you each person's voice separately. Tested on calls it had never heard: → nobody gets lost → no voices mixed → no invented voices → 99.3% cleanly separated You don't even have to tell it how many people are talking. All the data here 👇🏼1d ago27K1411490230.69%
YA PUEDES DARLE LA MISMA MEMORIA A TODAS TUS IAS. HOY MISMO. Sin vaults. Sin carpetas. Sin tutoriales de 40 minutos. → Entras en Brain → Conectas Slack, Notion, Drive o GitHub → Enchufas Claude, ChatGPT o tus agentes → Preguntas lo que quieras Y ya. Cada respuesta te enseña de qué documento sale. Y si cambias de modelo, la memoria se viene contigo. La mayoría usa la IA como a un desconocido al que hay que poner al día cada mañana. 🔖 Guárdalo, esto lo cambia todo.frozen29d ago27K368403802432.57%
ESTA EMPRESA QUIERE QUE HABLAR CON UNA IA SEA COMO HABLAR CON UNA PERSONA Se llama Tavus. No hace IAs a las que escribes. Hace IAs con las que hablas cara a cara. → Te ve → Lee tu tono → Sabe cuándo hablar Por dentro, tres modelos trabajan juntos: → Phoenix: la cara → Raven: lo que ve → Sparrow: cuándo hablar Primero aprendimos a escribirle al ordenador. Luego a hablarle. Ahora el ordenador aprende a mirarte.frozen21d ago27K1698180270.83%
YOU'VE BEEN CODING WITH AN AGENT THAT ONLY SEES ONE MACHINE Reason breaks that. → One session, many machines → Local, teammates and cloud → Keeps running in the cloud → Your whole team joins in The jump isn't that it reasons better. It's that it no longer depends on where you're sitting. Link below ;)frozen9d ago26K8520110260.54%
YOU'LL NEVER HAVE TO ASK TO "SPEAK TO A REPRESENTATIVE" AGAIN Cresta x Marshawn Lynch: → No more "press 1…" → The AI knows why you're calling → It fixes routine issues instantly → Humans step in with full context This is customer experience in Beast Mode ⬇️ https://t.co/LPbrl5bp11frozen29d ago26K77690120.40%
HOLLYWOOD NO LONGER NEEDS RETAKES. DRAMA changes the facial expression of anyone in a video that's already been shot. → Upload the video → Pick the emotion → Set the intensity → Generate Same face, same lighting, same camera. Only the performance changes. Link in the first reply :)frozen16d ago26K31540331581.75%
AI IS STILL BUILT FOR ONE PERSON AT A TIME. One user. One assistant. One chat. But work doesn't happen like that. @zenzap_chat puts people and AI agents in the same group chat: → Agents join your team chat → Connected to your tools → Humans approve key steps → Work gets done in-chat Create an agent in under 90 seconds. No code. Work is a team sport. AI should be too.frozen8d ago25K69212090.37%
ONE PERSON CAN NOW RUN A FULL CONTENT TEAM WITH GROK One bot per job. You just make the calls. → Research competitors while you sleep → Pick the one idea worth testing → Build the assets with consistent avatars → Auto-schedule everything in Postiz No camera. No editors. No team. 🔖 Save this. https://t.co/0oyRrJaf9Mfrozen28d ago25K27713211221.34%
Your lawyer charges more the longer they take. @arceuslegal changes that: → Price before you start → Review in hours, not days → Licensed attorneys → Flat fee May the firm that treats clients best win, not the one that takes the longest.frozen21d ago24K97560130.49%
GOODBYE TO BUILDING A SERVER FOR EVERY CLIENT Agent37 launches today on YC: → One computer per client → Saves their files → Wakes up on message → Near-zero cost when idle Pay per minute and up to 94% less, they say. https://t.co/4wYRUFVKHkfrozen23d ago24K1427190240.79%
NADIE EN TU GRUPO QUIERE LLAMAR AL RESTAURANTE. ESTA IA SÍ LO HARÁ Rene ya funciona dentro de tus chats de grupo. Añádelo como cualquier otro contacto, di su nombre y se encarga de todo a la vista de todos: → Hace llamadas reales → Reserva y compra cosas → Calcula quién debe qué → Comparte vuestro tiempo libre → Diseña la invitación → Crea la web del viaje → Recuerda las cosas a todos Cualquiera del chat puede aportar. Los inicios de sesión y los pagos van a tu DM, y al grupo solo llega el resultado. Sin app. Empezar no cuesta nada. Solo escríbele 👇🏼frozen10d ago23K18616240611.22%
THE HARDEST PART OF AI FILMMAKING ISN'T THE GENERATION. It's proving the work is yours. This studio tracks it as you create: → Characters, shots, takes → Every human edit recorded → IP issues flagged early → Rights records attached One-shot is the starting point. Post is where you become the author.frozen6d ago23K44850230.34%
NADIE ESTÁ HABLANDO DE ESTO Y ES DE LO MÁS FUERTE DEL AÑO Un laboratorio independiente acaba de meterse en el top 5 mundial de edición de imagen. Ranking de Artificial Analysis: → 3º Google, 1152 → 4º Krea 3 Ultra, 1151 → 5º Microsoft, 1151 Arriba solo hay gigantes. OpenAI va a valer más de 1,4 billones. Meta gasta hasta 135.000 millones este año. Krea ha levantado 80 millones en toda su historia. Y está a un punto de Google.2d ago23K17620100291.03%
THIS IS GOING TO SAVE SO MANY RELATIONSHIPSfrozen6d ago23K581180210.43%
THIS IS WHAT A SUPPORT TEAM LOOKS LIKE IN 2026 A customer reports that their export stops at 10,000 of 18,400 contacts. And instead of passing the ticket from tool to tool: → Support agent replies → Coding agent finds the bug → Docs agent updates the guide → CRM agent prepares the renewal All on the same ticket, with the same context: logs, billing, code and previous conversations. It's called Helpin. It's open source and you can self-host it.frozen14d ago23K15814260301.00%
ASÍ SE VE UN AGENTE CONSTRUYENDO UNA APP DE VERDAD Le pides a Eazo un generador de pósters y esto es lo que hace: → Monta la app entera → Revisa su propio código → Detecta el fallo solo → Añade el login → Preview en móvil, en directo Escribes "un tema, una frase, un mood" y te sale el póster. Le das a Publish y ya es tu app. #eazo #personalagent #qoni14d ago23K12717130330.84%
THE WORST PART OF SHIPPING A PROTOTYPE WAS NEVER BUILDING IT It was turning "the header feels off" into a prompt. Now reviewers comment on the exact element. → Comments stay pinned → Threads sync live → Batch them all → One run to fix The gap between "here's my feedback" and "it's fixed" just got a lot smaller.frozen15d ago23K9020100290.66%
claude code writes the bug, sazabi finds the bug, claude code fixes the bug. https://t.co/MQx4i0EwCEfrozen9d ago22K7413120230.54%
🚨 ¿Entiendes lo que va a sacar Myndlab el 6 de octubre? Describes una app y te la construye entera: frontend, backend y base de datos. Y el código es tuyo, en un contenedor, para llevarlo donde quieras. Aquí lo puedes ver ⬇️frozen21d ago22K1356120290.82%
YOU CAN NOW BUILD APPS INSIDE CLAUDE OR CHATGPT AND SHARE THEM LIKE A GOOGLE DOC Charming just launched it and I tried it. I asked Claude for an app to pick Friday's movie with my friends: → I add five movies → My friends vote → The winner shows up Nothing to download. No extra subscription.frozen15d ago22K13219160320.90%
GOOGLE TIENE UNA IA QUE PREDICE PATRONES EN CUALQUIER COSA, Y ES GRATIS Predecir ventas siempre ha sido cosa de un analista caro o de una corazonada. Acaba de salir TimesFM 3.0 y ya no solo mira tu histórico. También entiende lo que pasa alrededor. → Varias métricas a la vez → Entiende promos, precios, festivos → Sin entrenarlo tú → #1 en fev-bench y TIME → Corre en tu Mac → Integrado en Google Sheets Lo que antes era un equipo de data science ahora es un pip install. 32.4k stars en GitHub. Código abierto bajo Apache 2.0. Link abajo :)frozen11d ago22K369312914494.07%
junior analyst at 2am watching an open-source model align the slides he's been nudging pixel by pixel since 11pm https://t.co/WJyNENKI2nfrozen8d ago21K11512100210.74%
BUILDING OPEN MODELS IN EUROPE? This is for you. → Open models, open weights → Coding agents, runtimes → Inference, serving tools → Post-training, RL, evals Compute grants on Nebius Token Factory, for European projects advancing the open-source frontier. Show what you'd build. https://t.co/rKDM3dlM3Vfrozen4d ago21K1501590250.94%
EVERY TIME YOU PASTE SOMETHING INTO AN AI, IT LEAVES YOUR COMPANY Contracts, code, client data... They end up on a server you don't even know the location of. Today Frontier/1 launches to change that: → Your AI in your country → Machines just for you → Nobody can copy your model → A receipt for every response It's like having your own safe for AI, instead of leaving everything at someone else's house.frozen16d ago21K21515150321.32%
ESTO ACABA DE MATAR A LOS EDITORES DE VÍDEO Sueltas los clips grabados en una carpeta y Claude Code te devuelve el vídeo terminado. Grabas con fallos, silencios y muletillas. Te olvidas. Y al rato tienes el final.mp4 listo para subir. Se llama Video Use y ya tiene 29.3K stars en GitHub. Hace lo que te quita horas: → Quita los "eh" y "umm" → Corta los silencios → Arregla el color → Pone subtítulos → Crea las animaciones Y antes de dártelo revisa cada corte él solo. Es como tener un editor que trabaja mientras tú grabas el siguiente vídeo. Open source. Funciona con Claude Code y Codex. 🔖 Guárdalo, merece la pena.2d ago21K346542913763.90%
A really interesting direction for AI. Instead of optimizing models to generate more text, decision models focus on a different problem: evaluating a question and assigning calibrated probabilities to defined outcomes. Fast, open, and built end-to-end in Australia. Curious to see what gets built on top of this.frozen17d ago21K20213131271.24%
DEPLOYING JUST GOT BORING (IN A GOOD WAY) Synteq just launched Cloud Apps, their new product to get your project online without touching infra. The hard part was never building. It was getting it live. Now you just: → Push to GitHub → Run synteq deploy → Or drag and drop No server setup. No config headaches. Live in seconds. Made for solo devs, indie builders and small teams. Not enterprise infra.2d ago20K9714100220.71%
MIXAR JUST PUT 400+ AI MODELS INSIDE ONE 3D WORKSPACE Mixar gives you access to LLMs, image-to-3D, texturing models and native 3D tools from the same environment: → One workspace → 400+ models → Agents can work directly inside the scene → Multiple agents can work across different scenes → Built on a fork of Blender, not another plugin No need to stitch together separate AI tools, scripts and 3D software. Try it now.frozen20d ago20K13417130270.95%
ESTO HA REVOLUCIONADO LA INDUSTRIA INMOBILIARIA Convierte fotos de una casa en un recorrido 3D fotorrealista. Es Gaussian Splatting, de Inria y el Max Planck Institut. 23.5k stars. → Solo necesitas fotos, ni app ni VR → El resultado se navega desde el navegador → Se está usando en inmobiliaria, Airbnbs y museos Es una licencia de investigación. Para vender escaneos con esto necesitas hablar con Inria. Guárdalo 🔖frozen26d ago20K129231411131.40%
THE AGENT SIGNED INTO FIGMA AND NEVER SAW THE PASSWORD. That one line in the video is the whole product. → Agent asks to log in → The vault handles it → The agent gets access → Your password stays yours Delegation without exposure. https://t.co/QvidV62cJufrozen7d ago19K13714110200.94%
me watching an AI cold call people so i dont have to 🧍frozen21d ago19K116690180.77%
CÓMO CONVERTIR UN VÍDEO EN 9 POSTS CON CLAUDE Tutorial de 15 minutos. → Encuentra los mejores momentos. → Recorta y subtitula en vertical. → Publica en TikTok, Reels y Shorts. Vídeo largo → 9 posts → 3 redes 🔖 Guárdalo antes de perderlo. https://t.co/fSX21Impt5frozen17d ago18K419503502524.10%
TODOS ESTOS MÓVILES LOS ESTÁ CONTROLANDO UNA IA. PhoneUse acaba de darles móviles a los agentes de IA. Pueden: → Pedir un Uber → Investigar tendencias en TikTok → Navegar apps antiguas → Pasar la verificación en dos pasos → Reaccionar a tus notificaciones La IA ya podía usar ordenadores. Ahora también tiene móvil.1d ago18K1662880731.53%
OKARA ACABA DE LANZAR UN EQUIPO DE MARKETING QUE SE ACTIVA PEGANDO TU WEB Se llama Dots for marketing: → Pegas tu URL → Despliega varios agentes → Buscan tráfico y usuarios → Reddit, X, LinkedIn, SEO Lo que antes era una agencia, ahora es un enlace. https://t.co/CjSenRyp9Ffrozen23d ago17K18619100651.63%
🚨 ACABAN DE PONERLE PRECIO A LOS ERRORES DE LA IA Una startup de Stanford acaba de salir de stealth con $10M para hacer algo muy concreto: → Revisar cada agente de IA → Encontrar dónde puede fallar → Calcular cuánto te costaría ese fallo Y después asegurarte contra ese riesgo. El ejemplo que enseñan en su web: Una empresa le da a su agente de reembolsos permiso para pagar solo. → Transferencias bancarias → Sin límite de importe → Sin pasar por soporte Un cliente pide $12.500 y el agente los paga sin más. La herramienta lo detectó antes de que llegara a producción. Las grandes aseguradoras de EE.UU. están pidiendo poder excluir la IA de su funcionamiento, y en la ronda está Adeyemi Ajao, cofundador de Tuenti. Por qué te importa si trabajas con IA: Lo que no se puede medir no se puede asegurar. Y sin seguro, ninguna gran empresa va a dejar que una IA toque su dinero. Se llama Harbor, enlace abajo ⬇️frozen9d ago17K18518230411.57%
ALGUIEN CONSTRUYÓ UN SEMRUSH OPEN SOURCE Y LO REGALÓ. Las suites de SEO cuestan entre 100 y 500€ al mes. Pagas por 40 funciones y usas 4. OpenSEO va al revés: te quedas con los workflows que de verdad usas y pagas solo por los datos que consumes. → Keyword research y rank tracking → Backlinks y competidores → Auditorías y visibilidad en IA → Sin suscripción, pagas por uso → MCP y skills para agentes Lo interesante no es la UI. Es que Claude Code (o tu agente) puede hacer el análisis de keywords entero mientras tú haces otra cosa. 22.800 stars, 3.000 forks y licencia MIT. 🔖 Guárdalofrozen13d ago14K171283011792.91%
YOUR AI WRITES THE CODE. WHO TESTS IT? That's the question nobody asked while everyone raced to generate more code. More code means: → More changes → More broken flows → More bug reports → Same number of engineers Ship is one answer: an agent whose whole job is checking what the other agents ship.17d ago14K12720110391.44%
TU ORDENADOR ESTÁ HECHO PARA TI, ESTE ESTÁ HECHO PARA LA IA Acaba de lanzar un ordenador en la nube hecho para la IA, no para personas. → Lee webs como estructura → No solo capturas → Un único SDK → Mandas la tarea → Recibes el trabajo hecho Llevamos años haciendo la IA más lista, y ahora por fin tiene un ordenador donde trabajar. Esto es el futuro... Link abajo :)frozen13d ago10K13315150482.10%
🚨 CLAUDE YA PUEDE EDITAR TUS DOCUMENTOS DE GOOGLE Pero con una regla que lo cambia todo: tú apruebas cada edición antes de que se aplique. Nada de copiar el texto en el chat y pegar la respuesta de vuelta. Claude trabaja dentro del propio archivo. Así funciona: → Abres Docs, Sheets o Slides → Claude aparece en la barra lateral → Lee lo que tienes abierto → Le pides el cambio → Revisas y aceptas Le puedes pedir que resuma un contrato, que busque incoherencias o que redacte un informe, sin salir del documento. Y también funciona al revés: pegas el link de un archivo en Claude, o le pides uno nuevo, y se abre al lado del chat para editarlo juntos. Todo respetando los permisos que ya tienes en Google. Ya está en beta para todos los planes de pago. 🔖 Guárdatelo.frozen15d ago9.5K12026230482.28%
72% OF READMA'S DOCS TRAFFIC IS ALREADY AGENTS, NOT HUMANS and your api is still written for a dev to read on a monday morning. restless flips that: → npx restless init and you're in → every request leaves the api better for the next one → agent recovery: the fix ships inside the error → agents report what's missing and it turns into docs your api isn't the back door anymore. it's the product.frozen10d ago6.2K8014100201.99%
video editors after reading "30 second scenes in one pass" https://t.co/hOy5yQs6e41d ago6.2K451190231.42%
DECISION MODELS JUST STOPPED BEING YES OR NO. Span-02 answers with: → Choices → Scores → Numbers → Straight from screenshots Less prompt engineering. Less output parsing. And you can stop it anytime and still get its best answer.frozen7d ago6.2K1317100192.71%
ACABAN DE SACAR UN PHOTOSHOP GRATIS 70.000 personas ya lo han guardado en GitHub. Se llama PhotoCraft y es una alternativa a Photoshop que corre en tu propio ordenador. Y no viene solo. Tiene versión de casi toda la suite: → Illustrator → Premiere → Lightroom → After Effects → InDesign Sin suscripción. Sin pagar por cancelar. Mac, Windows, Linux y navegador. 🔖 Guárdalo.1d ago6.0K7413150522.58%
THE HARDEST PART OF AI IMAGE TOOLS ISN'T THE PROMPT It's knowing what you want before you've seen it. Creative work rarely starts that clear. It starts with a vague feeling. Envato's new Burst mode gets that: → One starting point → Six directions back → One AI credit → Pick what clicks You stop guessing the prompt. You start choosing.frozen15d ago5.6K8117110272.42%
AI AGENTS ALREADY HAVE THEIR OWN EMAIL. AND THEY'RE USING IT AgentMail just passed 1 billion emails processed. → Under 20M a year ago → 10M+ inboxes created → 50x in one year Agents are processing invoices, answering customers and signing up for services. Through email. https://t.co/zXrhjUM5731d ago4.6K711080262.48%
CLAUDE ACABA DE SACAR SU PROPIO EMPLEADO Se llama Claude Duet y no espera a que le digas qué hacer. Te escribe él cuando algo necesita tu atención. Te permite: → Resumen cada mañana → Trabajar en segundo plano → Reuniones y facturas → Más de 6.000 apps Le pides algo, cierras el móvil y lo sigue haciendo en su propio ordenador en la nube. De momento solo en EE. UU., para Pro y Max, con lista de espera. Juan Pe Navarro lo explica genial en este vídeo. 🔖 Guárdalo.7h ago3.8K8321170323.98%
IDEOGRAM ACABA DE HACER QUE CUALQUIERA PUEDA CREAR UN LOGO PROFESIONAL CON UNA FRASE No te da una imagen. Te da el archivo vectorial de verdad: → El texto sale bien escrito → Cada forma es editable → Escala sin pixelarse → Abre en Figma o Illustrator Y si ya tienes un logo en JPG borroso, lo subes y te lo devuelve en vector limpio. Cambias los colores y los degradados se adaptan solos. Guárdalo.5h ago1.9K6017210296.66%
OK, ESTO ME PARECE UNA LOCURA Ahora puedes generar imágenes y vídeos dentro de Claude. Sin salir del chat. Así de fácil: 1. Abres Claude 2. Vas a Ajustes → Conectores 3. Pegas el link de Sume 4. Le pides el vídeo Y ya. Claude te lo genera ahí mismo. Sume es una plataforma con los modelos top de imagen y vídeo en un solo sitio: → Seedance 2.0 y 2.5 → Veo 3.1 → Nano Banana Pro Y lo mejor: 0% de sobreprecio. Pagas lo que cuesta el modelo original, nada más. Sin suscripción. Generas lo que quieras y pagas solo lo que usas. No entiendo cómo esto no lo está usando todo el mundo todavía.4h ago1.6K82292504211.24%
UNA FOTO TUYA + UNA CANCIÓN = UN VÍDEO TUYO CANTÁNDOLA. Se llama P-Video-2-Pro Avatar, lo acaba de sacar Pruna AI y funciona así: → Subes un retrato → Acepta voz o canciones → Hasta 3 minutos → Sin cortes de plano → 10 idiomas por TTS Un minuto a 480p cuesta 1,20$, y estas 2 semanas va a mitad de precio. Te dejo el anuncio aquí abajo :)4h ago1.5K611570287.24%