🧠 IA, superinteligencia y el mundo que viene ⚡ Pruebo herramientas que parecen ciencia ficción 🚀 Emprendedor por instinto. Humano, por ahora.
rafael.business23@gmail.com
Joined April 2023
- Tweets40.8K
- Following204
- Followers109K
- Likes31K
Pinned Tweet
¡Oye, @grok, en caso de una Tercera Guerra Mundial, ¿cuáles serían los 5 países más seguros del mundo y por qué?
Rafa Gonzalez retweeted
We switched 10 people from iPhone to Blue. Weeks later, none of them want their iPhone back.
Introducing Blue: a phone you just talk to. Tell it what you need, and it gets done.
It has all the context, can use every single app, and even triages your notifications.
We built it for people who always have a little too much to do. The ones buried in messages, notifications, and to-dos.
If that's you, get yours today.
Backed by Y Combinator + 70 other VCs & angels.
heyblue.com
Rafa Gonzalez retweeted
La diferencia entre comprar un COCHE AL CONTADO y FINANCIARLO que mucha gente descubre demasiado tarde:
🔴 PERSONA A: COMPRA AL CONTADO
🚗 Precio del coche: 20.000 €.
💳 Lo paga de una vez.
💰 Coste total: 20.000 €.
✅ Deuda pendiente: 0 €.
🔵 PERSONA B: FINANCIA EL COCHE ↓
La mayoría de las pruebas de IA miden cómo responde un modelo ante problemas nuevos.
Pero casi nadie evalúa si mejora después de enfrentarse a problemas similares.
Y eso es precisamente lo que necesitamos de un agente de IA: que mejore con la experiencia.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
One product photo, tiny ad budget.
I upload it to Dreamina Marketing Studio and generate images and videos.
Try this workflow for client ads:
Dreamina: tinyurl.com/376c8r3x
CapCut PC: tinyurl.com/2v23rn22
#CapCutPC #Dreamina
🤝 Paid partnership
I select the narrated clips and tell CapCut Agent EditPilot the correct spelling, asking it to fix the repeated caption typo across the group.
One conversational request replaces correcting the same word over and over. My patience appreciates it.
#chatwithcapcutagent
1
1
132
On the timeline, I move captions clear of the product, add music and a subtle transition, then preview and export.
Everything gets one last spelling check.
Put this workflow to work turning product images into ads you can offer clients.
#CapCutDidThat
1
138
Rafa Gonzalez retweeted
People don't understand how insanely high IQ Newton was. By age 20, he had no training in math or algebra and in <3ys he studied Euclid, Descartes, Kepler, etc., then laid the groundwork for the laws of motion & gravity, invented calculus, and the theory of colors & light. All while working alone on a little farm outside town.
There's a great clip of Neil deGrasse Tyson explaining how there's no contest as to why Newton was the greatest scientific mind ever produced.
In art, there is no target or framework to go about creating something. Only da Vinci could've painted The Last Supper, Homer written The Iliad, or Beethoven composed the Fifth Symphony. No other human being could've produced those pieces of art. Creative work has no boundaries and can't be pinned down or analyzed.
Science is different. In science, you need to know what questions to ask because the work you can do is deterministic, in a sense. You try to fill a piece of the puzzle that already exists, and nature dictates whether something is true or not. In science, which threads you pull is just as important as the work itself, if not more.
Someone would've eventually done what Newton did because the pieces are predetermined. However, in 1,000s of years and likely 100s thereafter, only Newton figured it out. And it was because of the questions he was after + his extreme intelligence.
For ex, in the 1680s, Halley was trying to understand what motion an inverse-square law would produce. No one had the answer, so he consulted Newton, and Newton said: "An ellipse." Halley asked how he knew that, and Newton's reply was sth like: "Oh, I proved it years ago." The guy had used the math he'd developed to answer it.
At the end of Newton's book on Opticks, he includes his unresolved Queries and other topics that interested him. It's astounding because he framed questions in a way that gave away his intuitions. Not only were the questions 100-200 years ahead of his time, but his guesses would prove largely right.
He really was someone who thought extremely hard about everything and paired the thinking with rigorous experimentation.
“I do not know what I may appear to the world, but to myself I seem to have been only like a boy playing on the sea-shore, and diverting myself in now and then finding a smoother pebble or a prettier shell than ordinary, whilst the great ocean of truth lay all undiscovered before me.” - Newton
Rafa Gonzalez retweeted
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
Rafa Gonzalez retweeted
Eu pagaria mais por uma IA que lembrasse como gosto de fazer as coisas, sem precisar repetir as mesmas instruções a cada nova conversa.
Uma IA realmente útil deveria se adaptar às nossas preferências com o tempo.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
Rafa Gonzalez retweeted
The part of Generative Latent Prediction I’d steal for other models: the next state is something you can compare against the actual next frame. More of this please.
Introducing PAN 2: an open foundation world model for physical intelligence.
A language model predicts the next word. PAN 2 predicts the next world: what happens after each action, grounded in time, physics and space.
Towards AI that simulates “what if?” before it acts.
More details: ifm.ai/pan/
El fundador de Anthropic acaba de afirmar que la IA podría curar el cáncer en 10 años.
Pero el fundador de NovaAtom, @QuanquanGu, predice que solo harán falta 4.
La IA ya conquistó el lenguaje y la programación. La biología es la próxima frontera.
El capital ya está fluyendo hacia la IA aplicada a la ciencia: Isomorphic, valorada en unos 40.000 millones de dólares; 1.800 millones del Gobierno de EE. UU., Google, Meta y Biohub; y el primer fármaco diseñado con IA en fase III.
Ahora, un equipo de solo 15 personas ha lanzado NovaAtom-Lite y está superando a Isomorphic, la empresa de Demis Hassabis valorada en 40.000 millones de dólares.
> 79,07 % en FoldBench para anticuerpo-antígeno, frente al 75,58 % de IsoDDE.
> 77,06 % en proteína-ligando, frente al 75,99 % de IsoDDE.
La API ya está disponible.
🤝 Paid partnership
Introducing NovaAtom-Lite, an all-atom structure prediction model for drug discovery.
It predicts the structures of proteins, nucleic acids, and the molecules they bind, including how they fit together.
NovaAtom-Lite outperforms both IsoDDE and AlphaFold 3 on FoldBench antibody–antigen structure prediction (DockQ ≥ 0.23):
NovaAtom-Lite: 79.07% (n=172)
IsoDDE: 75.58%
AlphaFold 3: 47.90% (n=167)
Our evaluation uses 5 seeds × 5 samples, with the top-1 prediction selected by the model's own ranking score.
Better structure prediction. A stronger starting point for your next experiment.
Live today via Geodesic API and NovaDDE.
🔗 API: platform.geodesiclab.com/
🔗 NovaDDE: agent.geodesiclab.com
🔗 Model Card: geodesiclab.com/news/introdu…
Rafa Gonzalez retweeted
Introducing NovaAtom-Lite, an all-atom structure prediction model for drug discovery.
It predicts the structures of proteins, nucleic acids, and the molecules they bind, including how they fit together.
NovaAtom-Lite outperforms both IsoDDE and AlphaFold 3 on FoldBench antibody–antigen structure prediction (DockQ ≥ 0.23):
NovaAtom-Lite: 79.07% (n=172)
IsoDDE: 75.58%
AlphaFold 3: 47.90% (n=167)
Our evaluation uses 5 seeds × 5 samples, with the top-1 prediction selected by the model's own ranking score.
Better structure prediction. A stronger starting point for your next experiment.
Live today via Geodesic API and NovaDDE.
🔗 API: platform.geodesiclab.com/
🔗 NovaDDE: agent.geodesiclab.com
🔗 Model Card: geodesiclab.com/news/introdu…
Rafa Gonzalez retweeted
Pra mim, agente de IA tem que ficar mais rápido e mais barato quando faz a mesma tarefa pela décima vez. Se não fica, tem alguma coisa errada.
🤝 Paid partnership
OpenClaw, Hermes Agent and Prime Agent don't meaningfully learn from experience.
On Continual-ARC, our new benchmark at @letsunifyai, none of them were able to memorize anything of value and none were able to generalize to new tasks. Training a small 9B model's weights did both.
La diferencia entre comprar un COCHE AL CONTADO y FINANCIARLO que mucha gente descubre demasiado tarde:
🔴 PERSONA A: COMPRA AL CONTADO
🚗 Precio del coche: 20.000 €.
💳 Lo paga de una vez.
💰 Coste total: 20.000 €.
✅ Deuda pendiente: 0 €.
🔵 PERSONA B: FINANCIA EL COCHE ↓
El problema no es financiar un coche.
El problema es mirar solo la cuota y olvidarse del coste total.
Además, tener una cuota asumible no significa que puedas permitirte el coche.
Si haces lo mismo con el coche, el móvil y los muebles, puedes acabar cobrando para pagar decisiones pasadas.
La pregunta es sencilla:
¿Te comprarías ese coche si tuvieras que pagar hoy todo su precio?
2
11
26,023
