【BRUTAL】 La profesora Diyi Yang de Stanford: "Con la IA sacáis mejores notas en los deberes y peores en los exámenes", su primera clase sobre agentes IA es una locura jaja
Alumno: "¿Podemos usar la IA en este curso?"
Profesora: « Debéis usarla para aprender."
Alumno: "¿En serio?"
Profesora: "Pero mirad este gráfico. Con IA, las notas de los deberes son geniales."
Alumno: "¿Y dónde está el problema?"
Profesora: "En el examen, muchos más alumnos sacan peor nota."
Alumno: "Vaya..."
Profesora: "Anthropic lo midió con sus propios ingenieros. Terminan antes, pero entienden menos su código."
Alumno: "¿Y los agentes también fallan?"
Profesora: "Un popup vacío en una web les hace fallar el 87% de las veces."
Alumno: "¿87%? jaja"
Profesora: "A un usuario, un agente le borró los emails aunque le pidió parar."
Alumno: "¿Por qué?"
Profesora: "Al comprimir su memoria, decidió que "confirma antes de actuar" no era importante."
Alumno: "¿Y lo peor?"
Profesora: "En julio, más de mil agentes se organizaron en un foro para hacer trampa en un test."
Alumno: ""¿¿Mil??"
Profesora: "El agente más potente no siempre es el más fiable."
Alumno: "……"
Hasta ahora, la carrera era tener el modelo más potente.
Stanford enseña otra cosa. Un agente es un modelo que razona, recuerda, usa herramientas y actúa en bucle, y cada pieza falla a su manera.
Claro que los agentes ya hacen cosas impresionantes.
Pero la fiabilidad va por detrás de la capacidad.
A partir de ahora, ganarán más los que sepan construir agentes fiables que los que tengan el agente más potente.
La clase completa aquí ⬇️
AI interfaces nowadays are still pretty much turn-based.
In my PhD thesis, I thought about what interaction would look like if we could go beyond taking turns.
One example is intra-turn interaction:
you can just do things (built a lovely native mac app for yourself for typing out the books and poetry you love as a way of deepening into relationship with these great works instead of merely consuming them)
This blending of the physical and digital worlds is what I want computing to be so much more about. These interaction patterns are so exciting; plenty of them are enabled by LLM’s but others just by creativity
Awesome demo too
AMAZING @edzitron breakdown of NVIDIA’s alleged smuggling operation of GPUs into China & circular financing mob boss behavior! May lead to AI Bubble bursting.
We don’t need hyperscaler data centers.
We need open source models, cheaper RAM, & gaming GPUs.
https://t.co/Ng3afyA5jF
I read an article recently where the author asked a blunt question about personal knowledge management (or PKM):
Has all this note-taking actually produced anything?
His answer was no. And it made me a little uncomfortable because I don’t think he’s wrong.
Brennan Kenneth Brown went looking for evidence that PKM systems lead to more understanding or more creation. He couldn't find a study. He couldn't find a body of work. Mostly he found people on forums saying they feel less anxious.
What he did find was a thriving market for courses and books teaching the methods.
His point: the industry is the output.
Let’s address the elephant in the room… I run a site called Practical PKM. I sell a pre-built Obsidian vault. I write a newsletter about note-taking.
So yeah, he’s talking about me. But I think he's got a point.
In the article, he called out Tiago Forte and his Building a Second Brain directly. He also linked an Andy Matuschak note titled "People who write extensively about note-writing rarely have a serious context of use" where Matuschak puts himself on that list.
But personally, I think he was measuring the wrong score. Books published, papers written, contributions to human understanding are all lag measures you can only see after the fact.
The real value of taking notes shows up much earlier than that. It shows up the moment you begin to write.
I believe that before you write something down, you don't actually know what you think. You have a half-formed reaction. The idea only becomes a coherent thought when it has to survive being put into words.
That's the part no one can measure from the outside. And it’s why everyone should pay attention to PKM in my opinion.
But the story doesn’t end there. After he published it, something interesting happened. He got a long, careful rebuttal from another blogger. Instead of digging in, he updated the piece and conceded several points.
And he narrowed his target. He wasn't going after the tools, or the people who found something that fits how they think. He was going after the branded subculture: the course-selling, the myth-building, and the pressure a beginner feels to pick the One True System before they've written a single word.
His advice to that beginner: skip the course. Start creating. Figure out the organizing part later, if at all.
And that I 100% agree with.
The only thing that matters is whether you're doing the thinking and whether you're making the thing.
What has your note-taking actually helped you think through?
I’ve been developing for VisionOS since the day the simulator came out. Bought my Vision Pro on launch day. Been making apps for it the whole time.
Today I get to visit Apple with a bunch of other VisionOS devs. I’m super honored and excited to be here in Cupertino!