Fran Castillo (@francastillo) header
Fran Castillo (@francastillo) profile picture

AI is here — I help you turn it into business value.

Madrid-Almeria
Joined March 2009
Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)Photo by Fran Castillo (@francastillo)
Fran Castillo retweeted
Esto es básicamente lo que ocurre cuando le pides a la IA que te ayude a tomar una decisión.
in 2024, one our early investors came to Mercor’s office and sat down with all of our employees. he warned the entire team about why we should avoid the data market. our thesis was deeply contrarian at the time. VC advice is almost always market consensus, which erodes any alpha from being contrarian. if we’d taken the advice, we would have foregone tens of billions of dollars in value.
1
1
3
1,732
AI makes producing easy. It doesn’t make judging easy. It lowers the expertise needed to produce an answer, but not the expertise needed to recognize a good one. That’s where your value, and your team’s, now sits.
1
52
Where is AI heading? Even the people running the frontier labs are still figuring it out. In this edition of Coworkings AI, Carlos Almansa Ballesteros and I share where we’ve landed for now. Take a look 👇 coworkings.ai/p/coworkings-a…
26
Really exciting to see this new Agent-led growth motion in the market. Agents that help users move across the lifecycle: onboarding, activation, engagement and monetization.
2
49
Will AI replace people? Not yet. 🙂 Every useful AI agent we've deployed in two years needed a human deciding: How it sounds What it knows What it can and can't do When it escalates What good and bad look like When it goes live Which failures to fix first
1
55
AI is moving so fast that we are not sure this will hold for long. 😉 For now, an experienced teammate still makes many decisions, debates them, and getting them right after getting a few wrong first. coworkings.ai/p/coworkings-a…
1
33
Also in Coworkings AI: Jev, the model everyone's talking about, and why its value proposition is relevant for the market
20
Fran Castillo retweeted
The holy grail for robotics is being able to generalize: doing work in unseen places We rented 30 homes in the Bay Area and are doing tasks without any new training
599
749
458
7,899
2,351,677
How to pick the right LLM (Fable, Opus, Sonnet, ...) for your task 👇 The AI market is maturing & our understanding of it is improving fast. Most companies use LLMs on two main fronts: chatbots and agents. What is still unclear is which model to reach for depending on the goal.
2
2
96
4. How does the effort level influence the output? Is a lower-capability model at high effort better than a higher-capability model at low effort? Here we cover the main Claude models, with recommendations on the best model and effort level based on the difficulty of the problem
1
21
Fran Castillo retweeted
Has AI solved the Navier–Stokes problem? Not quite. The ideas came from mathematicians, while LLMs helped extend and verify them. AI may transform mathematical research, but the hype obscures the leading role of humans. (below my drawing, not AI) science.org/content/arti...
12
43
2
194
8,532
[1] Cada vez que leo "eso está generado con IA", me pregunto qué quiere decir: ¿Que todo el texto salió de un único prompt? ¿Que un texto escrito por una persona se editó con IA? ¿Que algunas partes se reescribieron con IA? ¿Que las ideas iniciales del post vinieron de la IA?
1
1
88
[2] Y si alguien trabaja con una ventana de contexto amplia y todo su histórico acumulado, y lanza un prompt, ¿eso está generado con IA o está generado con su propia información? No sé. No sé a qué nos referimos con "esto está generado con IA"
1
38
A few editions ago in "Coworkings AI" we looked at how AI can help coworking and flex spaces sell more, with three use cases from different stages of the customer lifecycle: acquisition, conversion and expansion.
1
1
68
This edition looks at a fourth, further along the lifecycle: cross-selling and personalisation through "intelligent" segmentation. coworkings.ai/p/coworkings-a…
1
50
It's an interesting use case where we combined traditional unsupervised machine learning models with LLMs.
16