marronald retweeted
Your agent doesn't need 115,000 tokens of context to answer one question
It needs the right 5,100
Living Brain: 22.6x fewer tokens, 92.2% accuracy on LoCoMo, 464ms retrieval
Bigger context windows were never the answer
🤝 Paid partnership
marronald retweeted
this is actually insane
they just made ai agents rewrite the foundations of modern math as code
16.5 million lines.
the entire stacks project, 7,500+ pages of algebraic geometry, the stuff behind fermat's last theorem
every single line checked by a computer. zero hand-waving
they ran the agents like a startup. some owned the architecture, some the definitions, some the proofs
and people still say vibecoding at scale = garbage
open source btw
🤝 Paid partnership
marronald retweeted
This is what fixing AI agents should look like
Your agent fails in prod, @neosigma_ai catches it, turns it into a test, and checks your fix against the real thing before you ship
No more evals from a spreadsheet nobody has updated in months
neosigma.ai
🤝 Paid partnership
The intelligence frontier keeps advancing every day. Are your agents delivering better outcomes on what matters?
For enterprises, the promise of agents is more work completed reliably, with less human intervention and a lower cost per successful task.
Today, we’re launching @neosigma_ai and our mission to empower every organization to build their own sovereign intelligence lab.
A place to operationalize their knowledge, replay real work, and optimize agents against the outcomes that matter to their business. Where production experience compounds into intelligence the organization owns.
We’re starting with NeoEval, our automated evaluation engine. It turns your production data into evaluation environments - that dynamically evolves along with your users, workflows, and frontier.
A playground for evaluating and optimizing your agents.
Your data. Your institutional knowledge. Now operationalized into evals and compounding intelligence.
Stay tuned for the next few days of updates from NeoSigma.
marronald retweeted
AI is about to start doing math faster than humans can check it
so axiom took the stacks project, 7,500+ pages that underpin fermat's last theorem, langlands and half of modern math, and turned it into 16.5 million lines of verified lean code
no shortcuts. every definition and theorem has to plug into thousands of others or the whole thing collapses
basically a golden record of math, written in the language agents actually speak
and it's open source
🤝 Paid partnership
marronald retweeted
> hand the agent a bug
> close the laptop
> take the dog out
> open your phone: it's already on step 4
Aether gives claude code its own cloud machine, so nothing depends on your laptop staying awake. you only pay for the machine while it's working
🤝 Paid partnership
Today we’re introducing Aether @aetherruntime.
The fastest cloud coding agents in the world that you can use with your existing subscription
Give it a task and just 20 seconds later your build is running on its own cloud machine. You can pick up from your phone, watch every line, and step in whenever you want as if it was local. Close the laptop. It keeps going.
It runs Claude Code or Codex on the Claude or ChatGPT plan you already pay for so there’s not a 2nd AI bill. You only pay for the machine.
When it’s done, if code review is enabled, a reviewer agent runs the code with the changes, which catches classes of bugs that can't be found from static analysis.
Before building Aether I built infra at Apple, Uber, and HeyGen. I built Aether because the cloud agents I tried before all had drawbacks, and I couldn't imagine building something without it now.
It’s free to start. runaether.dev
marronald retweeted
Everyone's using the same models now
The difference is going to be what your company does with its own data. Every time an agent fails and someone on your team fixes it, that's knowledge most companies just lose
@neosigma_ai keeps it and uses it to make your agents better at your business
Basically your own private AI lab
neosigma.ai
The intelligence frontier keeps advancing every day. Are your agents delivering better outcomes on what matters?
For enterprises, the promise of agents is more work completed reliably, with less human intervention and a lower cost per successful task.
Today, we’re launching @neosigma_ai and our mission to empower every organization to build their own sovereign intelligence lab.
A place to operationalize their knowledge, replay real work, and optimize agents against the outcomes that matter to their business. Where production experience compounds into intelligence the organization owns.
We’re starting with NeoEval, our automated evaluation engine. It turns your production data into evaluation environments - that dynamically evolves along with your users, workflows, and frontier.
A playground for evaluating and optimizing your agents.
Your data. Your institutional knowledge. Now operationalized into evals and compounding intelligence.
Stay tuned for the next few days of updates from NeoSigma.
marronald retweeted
Someone made a game about lying to AI.
You slip fake moves into the enemy fighter's head and watch it fall for them.
Free to play in your browser: jevfight.com
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!
marronald retweeted
Claude ya puede hacer 3D
Esto es Opus 5.5 + Tesseract renderizando una escena completa con ray tracing
Hay gente ganando +10K$/mes con canales de documentales sin mostrar la cara como este
Puedes empezar uno hoy mismo solo con Claude. y Tesseract es totalmente gratis
El workflow:
1. Pídele a Claude 10 ideas de documentales de tu nicho y quédate con la que tenga mejor gancho
2. Pide a Claude que te escriba el guion completo dividido en escenas
3. Para cada escena, Claude monta el plano 3D en Tesseract. movimientos de cámara, iluminación, texto animado
4. Mete tu voz en off y dile a Claude que sincronice los planos con ella (o pídele a Claude que locute el vídeo por ti)
5. Retoca lo que quieras con una frase. "cámara más lenta", "luz más cálida", "título más grande". solo cambia eso
De la idea al vídeo terminado sin abrir un editor
Opus 5.5 or Astra can now create professional-grade 3D outputs with path-traced rendering.
Using Tesseract with your agent, you can simulate how light moves through a 3D scene, bounces off surfaces, casts shadows, and creates reflections.
The video below was created without Blender or video models, and it remains fully editable.
marronald retweeted
This AI could change how chips are designed
It’s called Case, built by Polished Snow, which is coming out of stealth today.
Chip design and robotics companies are already using it in their daily work.
Designing chips takes years and huge teams. With demand for more computing power, speeding up that process could have a ripple effect across the entire tech industry.
Case works inside the systems these teams already use, handling everyday tasks so engineers can focus on the hard problems.
AI helping build the hardware it needs to grow.
Join the waitlist: plsn.ai
Humanity has always found ways to do more than it once could.
Our mission is to take human progress further by solving the existential constraints that remain.
plsn.ai
marronald retweeted
OKAY lo de esta empresa es realmente aterrador…
Humanity has always found ways to do more than it once could.
Our mission is to take human progress further by solving the existential constraints that remain.
plsn.ai
marronald retweeted
ESTO PODRÍA ACABAR CON EL MAYOR PROBLEMA DE LA IA
Polished Snow acaba de salir de stealth con Case, una IA que trabaja dentro de los sistemas de las empresas y se encarga del trabajo continuo del día a día, para que los ingenieros se centren en los problemas difíciles
Y ya la usan empresas que diseñan chips y robots
Ahora piensa en el problema más grande de la IA ahora mismo
No hay computación suficiente. Las GPUs están agotadas, los precios disparados y hasta las grandes empresas hacen cola para conseguirlas
Y diseñar un chip es lentísimo. Lleva años, cuesta una fortuna y solo unas pocas empresas en el mundo saben hacerlo
Si la IA empieza a ayudar a diseñar chips:
→ Salen chips mejores en menos tiempo
→ Cada chip rinde más, así que hay más computación con lo mismo
→ Más empresas pueden diseñar sus propios chips
→ Y a la larga, una IA más barata para todos
Es la IA ayudando a construir lo que necesita para crecer
Lista de acceso en plsn.ai/
Humanity has always found ways to do more than it once could.
Our mission is to take human progress further by solving the existential constraints that remain.
plsn.ai
marronald retweeted
Tell ChatGPT something. Open Claude. It already knows.
That's Living Brain, one memory that follows you across ChatGPT, Claude, Cursor, Codex, OpenClaw…
Switching models shouldn't mean starting from zero
🤝 Paid partnership
marronald retweeted
Esta es la única cuenta que necesitas seguir si quieres aprender sobre IA y estar al día
Todavía no le he visto un solo tweet malo
marronald retweeted
ESTO ES ORO PURO
Cómo crear tu primer agente de IA, paso a paso
Si hubiera tenido esto hace un año, habría sacado mi primer agente en un día y no en dos semanas
Guarda la imagen, pásasela a Claude y pídele que te monte el tuyo de principio a fin
marronald retweeted
Anthropic CEO after finding out Sage2 is 2.0–2.3× cheaper than Haiku 5.5 and makes decisions 3× faster
Software can now understand human language and act on it, in real time.
Introducing Sage2
- Extract: natural language in, structured data out. 166x cheaper than Opus 5.5, beats @Cloudflare's Clairic
- Decide: the first decision model with streaming. 3x faster, because it reads the context as it grows.
Sage2 is a new primitive for developers. Get 20M tokens free 👇
Also available on @near_ai and, for agents, via @mpp with @Stripe cards and @Tempo stables.
marronald retweeted
Finally a bank that cancels the subscription instead of just showing me I'm still paying for it😭
Meet Mal AI, the world’s first bank-native AI agent. It doesn’t just show you your money. It acts on it.
Here’s why we built it 🎥
@mal_ai_official
Join us!!!
marronald retweeted
Okay this is insane…
You and your friends can jump into the same world and change it just by typing in a chat
New map, new style, all live while you play, and everyone sees it at the same time
Explore Parallax at parallaxworld.live/ and follow @parallax_live for updates
Meet Parallax, the world's first real-time, live-editable multiplayer world model.
Share one world.
Reshape it with natural language.
See changes synced across player views.
Parallax — Worlds Without Limits
Explore Parallax: parallaxworld.live/
Follow @parallax_live for updates.
marronald retweeted
Como alguien que edita contenido para redes, sé que una buena idea no siempre se convierte automáticamente en un buen video.
A veces el resultado generado necesita pequeños cambios para que el ritmo funcione.
Este anuncio del Día de Muertos es un buen punto de partida: primero muestras el producto y los detalles festivos, después introduces las instrucciones en CapCut PC y finalmente obtienes una propuesta de video promocional.
Con las herramientas de edición de la línea de tiempo puedes recortar tomas, reorganizar escenas y mejorar el ritmo.
Porque generar el video es solo el comienzo; los últimos ajustes también cuentan.
Puedes explorar este flujo de trabajo aquí:
tinyurl.com/2v23rn22
#capcutpc #capcutmarketing #capcutblackfriday
🤝 Paid partnership
marronald retweeted
"Book me a flight to Tokyo, keep it under budget."
Meet Mal AI, the world’s first bank-native AI agent. It doesn’t just show you your money. It acts on it.
Here’s why we built it 🎥
@mal_ai_official
Join us!!!
marronald retweeted
I have a folder full of raw takes I never got around to editing.
OpusClip 4.0 → drop the file, switch on captions, speech cleanup and a viral hook, get back a finished video.
That folder is about to get a lot smaller.
🤝 Paid partnership
OpusClip 4.0 is here.
We used to find the best moments in your videos. Now OpusClip edits the whole thing, from the raw take to every format you post.
Turn on the edits you want, save your setup as a Recipe, and post everywhere. 🧵
opus.pro