CLOSED retweeted
Generating a 10 second video used to mean ordering a coffee and waiting.
Flux 3 Max does the opposite:
→ A 10 second video in just 12 seconds
→ According to Arcads, 22x faster than Seedance 2.5, with the same quality
→ Already available on Arcads
And when iterating is this fast, it changes how you make ads
Flux 3 Max is insane:
> 22X Faster than Seedance 2.5
> Similar quality
Live on arcads.ai today
CLOSED retweeted
you have no idea how much easier it gets with Reasons as your teammate
instead of opening 15 chats, here's what it gives you:
- powered by your own key (BYOK) or your subscription
- cloud or local, your choice
- shared with your team
- all in one app
🤝 Paid partnership
Introducing the best Grok Bot Alternative
Meet Reason Machines
With Reasons as your teammate powered by BYOK/Sub you can code much more!
Free trial, cloud/local, and shared all in one app.
Try it for free today at reasonmachines.com
CLOSED retweeted
AHORA PUEDES APUNTAR A CLAUDE CODE A CUALQUIER WEB, APP O JUEGO Y CONSEGUIR TU PROPIA VERSION
Craft Anything acaba de pasar a open source: 314 herramientas, Rust puro, más de 1.600 estrellas en GitHub desde el sábado.
funciona como una sala limpia:
paso 1 → estudiar. mira el original como lo haría un visitante: capturas a todos los anchos, texto visible, maquetación, colores, fuentes, espaciado
paso 2 → construir. tu propio Claude Code escribe un proyecto totalmente nuevo a partir de esas notas, con el stack que le pidas
paso 3 → comprobar. compara tu versión con el original en 3 anchos de pantalla y lista lo que no coincide
paso 4 → pulir. arregla las diferencias y vuelve a comprobar hasta que se parezca mucho
lo que la mayoría pasará por alto:
el modo clean room viene por defecto. nunca lee el código, los archivos ni el código fuente de la página original. las herramientas que intentan mirar dentro se bloquean.
un solo comando ejecuta todo el ciclo:
craft-anything-cli rebuild example.com “Recreate this website”
también reconstruye juegos en Unity, Unreal o Godot, y funciona como servidor MCP para cualquier agente.
está en alpha temprana, así que espera algún fallo. úsalo con software que sea tuyo o que tengas permiso para analizar.
CLOSED retweeted
What stands out to me most about Higgsfield Miracle is that you can do post-production while you're watching the video.
It turns the video into layers and you edit it piece by piece in real time, without waiting for it to render.
Coming soon to Higgsfield MCP.
Introducing Higgsfield Miracle.
Our latest proprietary post-production technology vectorizes any video into layers for localized, pixel-level editing in real time, with world knowledge built in. Every frame holds the image depth needed for cinema-grade post-production.
Now in private beta with select testers. Coming soon to Higgsfield MCP.
Set up Higgsfield MCP in Claude to be the first to try it.
CLOSED retweeted
edición de vídeo en tiempo real + agentes de IA es una combinación que quiero ver
en vez de regenerar todo el vídeo, cambias solo la parte que falla:
> una capa cada vez
> en tiempo real
> con la profundidad de cada fotograma ya ahí
Higgsfield Miracle convierte un vídeo en capas para editarlo por zonas, y llegará a Higgsfield MCP. Configúralo en Claude para ser de los primeros
para los creadores de anuncios con IA, eso podría significar muchas menos regeneraciones completas
🤝 Paid partnership
Introducing Higgsfield Miracle.
Our latest proprietary post-production technology vectorizes any video into layers for localized, pixel-level editing in real time, with world knowledge built in. Every frame holds the image depth needed for cinema-grade post-production.
Now in private beta with select testers. Coming soon to Higgsfield MCP.
Set up Higgsfield MCP in Claude to be the first to try it.
CLOSED retweeted
In CAD, a good-looking design is not a result.
Archimedes splits the job in two:
→ AI explores the design options
→ Reynolds, an open-source agent, runs the physics
They were able to cut the cooling costs of an entire data center by over 200k!
And pricing is results-based, not hourly.
🤝 Paid partnership
CLOSED retweeted
Anthropic sacó este plugin desactivado. Su nombre lo dice todo: “You should know”
1- Actualiza Claude Code a la v2.1.287 o superior (claude update) e inicia sesión con claude.ai o con una key de Console
2- Abre una sesión y escribe: /plugin enable cc-plugin-you-should-know@builtin (¿te dice que no está instalado? prueba en tu siguiente sesión)
3- Dale a Claude una tarea más larga y sigue trabajando como siempre
4- Déjalo activado: un agente secundario relee el chat de vez en cuando, con el mismo modelo, y esas revisiones cuentan para los límites de tu plan o tu factura de API
5- Fíjate en la nota que aparece encima de tu prompt cuando detecta una decisión que Claude tomó sin avisar, o un resultado que podría estar mal
6- Escribe 1 en un prompt vacío para saber más, 2 si ya lo sabías, 0 para descartarla
7- Desactívalo cuando quieras: /plugin disable cc-plugin-you-should-know@builtin
El agente secundario no tiene herramientas: no puede leer archivos, ejecutar comandos ni dirigir a Claude, solo deja una nota
CLOSED retweeted
AI agents can now rent their own iPhones.
Most agents live inside a chat window. does the opposite:
→ Its own phone: iMessage, TikTok, X, Uber, DoorDash
→ No API needed, it taps and types like a person
→ iPhones and Macs on rent, whenever it needs one
wait, should they be able to do this?
🤝 Paid partnership
Give your agents their own iPhone and Mac.
Al agents are smart enough to do almost anything, but they've been missing the one tool every human uses:
A device.
Introducing Glasshand.
Glasshand lets your agent tap, type, and swipe on real iPhones and Macs, exactly like a human.
That means it can:
→ Post TikToks
→ Book Uber rides
→ Send iMessages
→ Download any app it needs
If a human can do it on a device, your agent can too.
Rent an iPhone and a Mac for your agent at glasshand.com
CLOSED retweeted
AI agents can now have their own phones.
iMessage, TikTok, X, Uber, DoorDash. anything you can do on a phone, it can do now.
wait, should they be able to do this?
Give your agents their own iPhone and Mac.
Al agents are smart enough to do almost anything, but they've been missing the one tool every human uses:
A device.
Introducing Glasshand.
Glasshand lets your agent tap, type, and swipe on real iPhones and Macs, exactly like a human.
That means it can:
→ Post TikToks
→ Book Uber rides
→ Send iMessages
→ Download any app it needs
If a human can do it on a device, your agent can too.
Rent an iPhone and a Mac for your agent at glasshand.com
CLOSED retweeted
If you're an engineer and want to bring code generation to the rest of the company (IT, HR, finance, legal), this one's for you.
Serval is building Catalyst: you describe a workflow in plain language and it becomes code your team reviews, tests, and ships. And eligible new engineers start with a paid luxury vacation week, worth up to $10,000.
Open roles: Core Engineering, Deployed Software Engineering, Deployment Strategist, Deployment Success Engineering, and Automation Engineering. @getserval @AlexAtServal
🤝 Paid partnership
Starting today, every engineer who joins Serval begins with a weeklong, $10,000 paid vacation. We’re building the next generational company. We need generational talent to see it through.
Our mission has always been to empower people across every team to turn their understanding of a problem into working, trustworthy automations. Making that function inside the largest enterprises takes world-class engineering.
That’s what you’d be joining us to do. There are foundational decisions still to make, and the engineers making them will shape the future of how big we can build this company.
We want you to arrive rested and ready to do the best work of your life. So your first week is a chance to get away.
And if a @getserval employee refers you, both you and that employee receive $30,000, on top of the vacation.
Come build with us: serval.com/careers
CLOSED retweeted
A lot of teams skip the documents they can't afford to parse.
r-1 mini from @reductoai does the opposite:
→ $1 per 1,000 pages
→ A 100-page PDF costs 10 cents
→ A million pages is $1,000
So the archive you've been leaving out is back on the table
CLOSED retweeted
the first thing i do with any document parser is throw a scanned financial table at it. if the numbers don't stay in their rows, nothing else matters
@reductoai just put out r-1 mini: $1 per 1,000 pages. a 100-page PDF is 10 cents
going to throw my worst scan at it
🤝 Paid partnership
CLOSED retweeted
Most video generation benchmarks score a finished clip. Interactive streaming video generation is a different problem: the model has to respond as the stream goes on.
StreamEval is a benchmark built for exactly that, and its leaderboard is live at streameval.ai.
The leaderboard ranks systems across three categories: Avatar Interaction, World Generation and Real-Time Video Editing. Together they let you compare models side by side instead of relying on a single score or a demo reel.
What I like most is that it's open-source and reproducible. The code is on GitHub, the datasets are on Hugging Face and ModelScope, and the paper is on arXiv, so you can rerun the evaluation yourself and check the results instead of taking them on trust.
The project brings together researchers affiliated with Tsinghua University, NVIDIA, Carnegie Mellon University, UC Berkeley, Princeton University, the National University of Singapore and Reactor.
For AI companies and model developers: you can submit an API endpoint to have your model considered for evaluation under the same standardized framework. That gives you a standardized, reproducible way to see how your system performs and how it compares with others.
If you work on video generation, multimodal models, digital humans or avatars, or you follow this space as a founder or investor, it's worth a look. Explore the leaderboard, read the code and test your own model.
Links in the comment 👇
🤝 Paid partnership
CLOSED retweeted
having someone to hand work to used to be a corner office perk.
now anyone with an email address can have an AI employee, free to start and no sign up. It works from Gmail or Outlook, with no new app 4.4B people use email.
@viktor_com just email [email protected] and connect your calendar.
🤝 Paid partnership
Starting today, anyone with an email address can have their own AI employee. Free to start.
@viktor_com started in Slack. Now Viktor has an email address, so billions of people can reach it.
Just email [email protected] to get started.
Put Viktor on cc to book a meeting. Or forward any kind of work.
CLOSED retweeted
Most AI image models treat a design like one flat picture; ask for one change and it redraws the whole thing.
Brush does the opposite:
→ understands how each element relates to the rest
→ plans the edit before making it
→ restyles, adapts to new formats, updates every page
Canva's second in-house model in a week. Craft creates, Brush refines.
🤝 Paid partnership
CLOSED retweeted
Most image models go straight from prompt to pixels. That's why the text often comes out warped or misspelled.
Craft, Canva's new image model, does the opposite:
→ A planner first writes out the composition: which elements appear, where they go, what matters most and the exact words
→ A second model then paints it at up to 2K across 30+ aspect ratios
→ Every headline and label is a planned element, built to come out legible and in the right place
Plans before it paints
@canva
🤝 Paid partnership
Introducing Canva Craft 🚀
Extraordinary image generation is now available to everyone, powered by our first frontier model built specifically for design.
With Craft, Canva is now the world's second most advanced AI lab for image generation on @DesignArena's global leaderboard.
See what it can do 👇
CLOSED retweeted
no os miento cuando os digo que quizá nos encontramos ante la mayor cuenta de Twitter de la historia relacionada con la IA, no he visto un tweet malo de esta cuenta
ESTO ES UNA MALDITA LOCURA
13 repos gratis que convierten una carpeta de notas en un segundo cerebro que un agente puede usar de verdad: captura lo que lees, busca en todo lo que has guardado y nunca empieza de cero
[01] capturar
obsidian-clipper: el web clipper oficial de Obsidian, guarda cualquier página como Markdown github.com/obsidianmd/obsidi…
markitdown: convierte PDF, Word, PowerPoint, Excel y audio en Markdown github.com/microsoft/markitd…
yt-dlp: saca los subtítulos de los vídeos sin descargarlos github.com/yt-dlp/yt-dlp
paperless-ngx: OCR y búsqueda de texto completo en todos tus escaneos y PDFs github.com/paperless-ngx/pap…
[02] estructurar
obsidian-dataview: convierte tu vault en una base de datos que puedes consultar github.com/blacksmithgu/obsi…
obsidian-git: historial de versiones y copia de seguridad automática de cada nota github.com/Vinzent03/obsidia…
[03] buscar
qmd: búsqueda local en tus notas, combina palabras clave, vectores y re-ranking, y tiene servidor MCP github.com/tobi/qmd
meilisearch: búsqueda mientras escribes, tolerante a erratas y con modo semántico híbrido github.com/meilisearch/meili…
[04] memoria del agente
letta: agentes con estado que aprenden con el tiempo github.com/letta-ai/letta
langmem: hechos, comportamiento e interacciones pasadas para la memoria a largo plazo del agente github.com/langchain-ai/lang…
MemOS: una capa de memoria para agentes, guardada como un grafo que puedes inspeccionar y editar github.com/MemTensor/MemOS
[05] que sea tuyo
quartz: publica tus notas en Markdown como una web github.com/jackyzha0/quartz
syncthing: sincroniza tu vault entre dispositivos, P2P y cifrado github.com/syncthing/syncthi…
Un montón de notas solo se convierte en un segundo cerebro cuando un agente puede buscar en ellas y recordarlas.
Todos los repos de arriba son gratis, y casi todo funciona en tu propio ordenador.
CLOSED retweeted
In a fighting game, a combo is a sequence of buttons.
Mental Kombat does the opposite:
→ Two AIs fight in a pit, and you train one with a strategy in plain words
→ Your attack is a "fake move": a line you slip to the rival AI
→ Their AI can't tell your lie from the truth, and does something glorious and dumb
🤝 Paid partnership
I've built jevfight.com - fighting game where both players are controlled by Jev models! 🥊
Tell me your fighter name and I will attack you!