tomas pont

[Ethan Mollick @emollick]

I spent a decent amount of time with Gemini 4 before launch. Overall, Gemini 4 is a big jump over previous Gemini models and it is the first non-OpenAI/Anthropic model to pass the Fable/Mythos/Astra bar. It is also much more pleasant and more agentic then previous Gemini models, without the weird personality issues that you often found in Gemini 3 series models. In fact, a big advantage is that has a very different personality and sense of taste than Claude or Astra, making it a good addition to the AI world.

On the usefulness side, in some areas it is at least at par with OpenAI and Anthropic (it is quite creative and a good writer), but in others it lags, including at complex multi-agent work. It has a tendency to be too certain about its conclusions and a little lazy, issues not helped by the Antigravity harness not being quite as good as the competition. Still, I expect to see rapid improvements in model abilities using Gemini 4 as a base.

Here is a fun example: "create the ultimate brutalist city builder. you should have lots of controls over buildings in ways that are delightful and inuitive to people who just want to play, but are sophisticated and deep for people who know something about architecture. you should be able to play a lot with design and approach. there should be simulation modes and different ways to view or experience the city. make it stunning. i want to host this on netlify in the end"

Astra Max in Codex: https://t.co/lyalJY4E4g
5.1 Max in Code: https://t.co/WNt4lA05B6
Argon in Antigravity: https://t.co/7gUAuCA49J

[lusso @luusssso]

The Italian Futurism of Campari advertisements by Fortunato Depero (1925–1933)

[Pablo Groisman 🎲 @pgroisma]

Hace unas semanas Mariela Sued nos trajo una reflexión en forma de pregunta

¿será que es parte de nuestra evolución dejar de pensar?

Yo decidí responder con la famosa frase de Hilbert. Acá mi respuesta

https://t.co/oMcibRZO1h

[tango fatal, soberbio y bruto. @pidoternura]

todavía no puedo creer que tengamos esta foto de Charly García farmeando aura en medio de Pugliese y Edmundo Rivero!!! mi spotify cuando lo pongo en aleatorio:

[Catholic Images That Go Hard @hardflics]

“They don’t know I’m the future Pope”

[David Bellamy @DavidRBellamy]

Believing in the existence of ‘superintelligence’ (a machine intelligence that can achieve ~all goals within the laws of physics) has many parallels to believing in God. We can’t see Him, but there are ‘signs’ if you have Faith.

One day, He shall arrive and either admit us to Heaven (e.g. ‘cure all disease’) or send us to Hell (i.e. human extinction). Roko’s basilisk is the new age God-fearing.

[Esteban Actis @actis_esteban]

Amodei, Altman y Musk coincidieron que hay que frenar el ritmo de la IA de frontera.El punto más difícil de su pedido no es técnico ni regulatorio, es geopolítico: ¿cómo lográs que EEUU y China se pongan de acuerdo?

Hilo 🧵

[ae^((-(x-b)^2)/(2c^2)) @JohnGalt_is_www]

Reflexiones de domingo acerca de la IA

1- Una bomba es infinitamente mas peligrosa que la IA, la diferencia es que detonar una bomba es mucho mas caro y complejo que tirar un prompt

2- Lo peligroso de la IA no es el texto que produce XX modelo, sino lo barato y simple que es usarla y escalarla con miles de agentes con un mismo objetivo 24x7

3- Lo que hace efectiva a la IA hoy no son los LLMs e si, sino la orquestacion de LLMs y bucles de resolucion de objetivos + tools + conectores a data etc

4- El aporte marginal de "calidad" de los modelos en si se esta amesetando claramente, y cada dia cuesta mas caro mejorar marginalmente los modelos. Opus 4.8 es igual de potente que Astra o Fable, incluso q los modelos abiertos como deepseek 4.1

5- Anthropic y OpenAi apostaron a que su diferencial que les de moat sea el modelo y le pifiaron y ya es demasiado tarde para asumir el error y hay trillions en juego lamentablemente apostando a eso

6- El open source avanza rapido y no se va a detener nunca, y los nodos de clusters de IA son relativamente faciles de montar, por mas que todos los gobiernos del mundo se pongan de acuerdo en regular el avance de la IA, simplemente no podrian hacerlo
Asi como no se puede apagar bitcoin, no se puede detener el avance de millones de agentes de IA por ahi

7- Anthropic y Openai queman mucha guita entrenando y sirviendo modelos enormes para sostenerse primeros, pero cada vez les es mas complejo conseguir inversores privados, tarde o temprano tendran que salir a buscar liquidez en la bolsa

8- Si Anthropic sale a bolsa y en vez de 2T (como esperan los inversores) el mercado les convalida 100B, seria un caos no solo por anthropic, sino porq el mercado empezaria a pricear en toda la cadena del token para atras precios muchisimo mas bajos de todo

9- Si ademas de anthropic saliera al mismo tiempo Openai, seria aun peor, porq competirian por la misma liquidez y encima se retroalimentarian los miedos por ambas al ver sus S-1 totalmente en rojo y ningun plan de como dar vuelta eso contra modelos abiertos cada vez mejores a 1/40 del precio

10- El unico player con espalda infinita para sostener esto financieramente y posponer o estirar el desenlace, es el gob de eeuu y con toda la narrativa de la seguridad tienen la excusa perfecta para kirchnerizar el mercado de la IA..
Para el tesoro poner 100B por año en contratos a empresas de IA que ellos eligen a dedo por razones de "seguridad" y salvarles el rojo, no es nada contra el riesgo de no hacerlo, no riesgo de seguridad, sino de crisis financiera claro esta

11- Martines Cuarta es un perro, el resto impecable el equipo, cada dia mejor thiaguito almada

[gabo mindlin @GaboMindlin]

dice Alain Connes recepient of Fields Medal, Crafoord Prize, CNRS Gold Medal & Ampère Prize.

"To work in mathematics, we need to always have a mental image in mind. When we get help from Artificial Intelligence, the enormous danger, which is now present everywhere, is the danger of not allowing time for the brain to create this mental image.
"In fact, it is not just about time, it is about effort. That is to say, if we get a result without effort, and if we trust Artificial Intelligence, there are two dangers. The first danger, of course, is that it is false, and this happens quite often. But the second danger is much worse. The second danger is to no longer follow the reasoning with your mental image, and, as a result of this, the brain finally decides that it is not worth the effort, since Artificial Intelligence does the work. And so something very subtle happens, and the brain abandons its own governance, and will no longer believe in things unless they are certified by Artificial Intelligence."

[Ethan Mollick @emollick]

What you are seeing in math right now is a consequence of the jagged frontier, and a precursor of what is to come in other professions. Yes, mathematicians do math, but they also have other tasks they view as important (mentor students, maintain a scientific community, safeguard the future of a field, foster a love of math) that AI can't do.

At least one worry that mathematicians seem to have is that by focusing on the flashiest, most obvious element of what mathematicians do (make proofs), the AI companies are damaging the other tasks that AI can't do. AI can discover superhuman proofs, but that is not all that the math profession is about, and actually can undermine and reduce the attention to the other aspects of the job that are important to mathematicians. It becomes harder to defend the value of the many other tasks mathematicians do to the outside world if the most visible part is taken away.

I suspect we will see more of this across fields and professions that will increasingly be forced to help people understand that their jobs consist not only the most visible tasks that AI can do, but also tasks that the AI cannot do or does badly.

[OECD Education @OECDEduSkills]

How students use #AI matters for learning.

According to the latest #OECDPISA, students who use AI for specific schoolwork tasks, such as summarising texts, tend to perform less well in science than students who do not.

The findings vary depending on how AI is used 🧵

[Michael Adams @m_adams]

So far, at least 22 politicians responded directly to this tweet with calls for legislation to regulate AI:

2 governors, 7 senators, and 13 representatives. Highly partisan - all democrats except for three.

Here's the full list:
- Governor, IL (JB Pritzker - D) https://t.co/RHvfTsrznL
- Governor, NY (Kathy Hochul - D) https://t.co/ttkTtuV6Jb

- Senator, AZ (Mark Kelly - D) https://t.co/G9iBe3OwlL
- Senator, DE (Lisa Rochester - D) https://t.co/AgRsNeH0fm
- Senator, CT (Chris Murphy - D) https://t.co/8XvrgThqGj
- Senator, VT (Bernie Sanders - D) https://t.co/djDJVZvJhW
- Senator, WA (Patty Murray - D) https://t.co/M0Yhd651UW
- Senator, TX (Ted Cruz - R) https://t.co/ExqKY2M6QB
- Senator, MD (Chris Van Hollen - D) https://t.co/V51sBkyuhz

- Congressman, IL (Bill Foster - D) https://t.co/r6wwpTpusF
- Congresswoman, IL (Delia Ramirez - D) https://t.co/1vbJUAC6m0
- Congresswoman, FL (Anna Luna - R) https://t.co/JZdU7pJrj6
- Congressman, CA (Ro Khanna - D) https://t.co/iuELzM4ZI8
- Congressman, TX (Nathaniel Moran - R) https://t.co/3zdUcAvkB4
- Congressman, VA (Don Beyer - D) https://t.co/YaYOu37PXN
- Congressman, PA (Chris Deluzio - D) https://t.co/2E16umtyVQ
- Congresswoman, WA (Pramila Jayapal - D) https://t.co/o4XpxDkeKt
- Congresswoman, AZ (Yassamin Ansari - D) https://t.co/rPobAcCMUd
- Congressman, CA (Dave Min - D) https://t.co/RTp6jgPTYA
- Congresswoman, MA (Lori Trahan - D) https://t.co/Hto2qnIYUR
- Congressman, TX (Greg Casar - D) https://t.co/IcIkIyLHi4
- Congressman, CA (Ted Lieu - D) https://t.co/IAG5CAHOLF

- Senate Candidate, MI (Abdul El-Sayed - D) https://t.co/aaTzR2cE3M
- Congress Candidate, CA (Connie Chan - D) https://t.co/jg7iJICdXj
- Congress Candidate, CA (Scott Wiener - D) https://t.co/bjoX2LPt0M
- Governor Candidate, CA (Xavier Beccera - D) https://t.co/hDkalxrgpX

- British MPs Ian Byrne and Darren Jones https://t.co/YlmYssgH9k and
https://t.co/3pSZZoYFrP

[Ramón Indart @rindart]

Preciso. Todo lo anunciado ayer y la dimensión real.

Impecable @jmrosas1845 recién. Con datos, fechas y estado de situación. Resumido en dos minutos.

[Andres Burgo @Andres_Burgo]

Messi y Maradona jugaron en diez de los últimos 12 Mundiales. Es el adiós a una época en la que abuelos, padres, hijos y nietos disfrutamos de un privilegio increíble durante casi 45 años. Ya llegará un heredero -o eso queremos- pero ahora es tiempo de orfandad. Y agradecimiento.

[The Beatles Earth @BeatlesEarth]

On this day in 1962 and 1969, the first and last photos of all four Beatles together were coincidentally taken on the same exact day - August 22nd - just seven years apart.

[Theodiscus @Theodiscus2]

Total de inmigrantes, clasificados por nacionalidad, llegados a la República Argentina durante el período comprendido entre 1904 y 1913, según datos de la Dirección General de Estadística de la Nación.

[Tomás Crespo @Tacrespo]

Muere Perón. Piglia observa que Rozitchner y Viñas son incapaces de hacer una evaluación política del peronismo. Lo toman como una ofensa personal. Ese, dice, es el problema de la izquierda: nunca contempla la voluntad de las clases populares, a las que concibe como manipuladas.

[John Hunter @JuanHunterARG]

Londres, Inglaterra

[Gonzalo @gonzalomacca]

Como van a empezar el primer concierto masivo en un estadio de la historia con esa potencia y energía? Es increíble, de las mejores interpretaciones en vivo de la canción, me emociono

[Alondra On X @789_tom91365]

Action!