Engineer at @polar_sh. Builder in public to the void. Catalan and Romanian living in Barcelona. Occasionally serious.
Barcelona
- Tweets 2,317
- Following 628
- Followers 4,732
- Likes 2,314
Muy buen artículo.
Empecé en una startup. Éramos unos 8, pero sacábamos adelante el trabajo de un equipo de 20, y todo eso sin IA.
Siempre quise montar algo por mi cuenta. Lo probé y no salió bien y acabé buscando trabajo, pero fue una experiencia de la que aprendí muchísimo: descubrí qué me gusta y qué no, y me acostumbré a salir de mi zona de confort. Con el tiempo, la recuerdo como una experiencia brutal.
Después pasé por empresas más grandes, incluso Amazon. En todas aprendí, pero para mí, donde más se aprende es en una startup. Y de lejos.
Tras esa etapa empecé a trabajar en @polar_sh. Cuando llegué, éramos solo 3 personas; ahora ya somos 18.
Es emocionante ver cómo crece algo desde sus primeros usuarios: cada vez llega más gente, hay que escalar el sistema, sumar personas porque el equipo ya no da abasto… y, aun así, mantener la energía y la ilusión. Es una sensación increíble.
A quienes estén pensando en montar algo: estamos en una época de transformación y creo que es uno de los mejores momentos para emprender y crear algo increíble.
6
196
This has been funny. Try to type it in ChatGPT or your AI.
Based on everything you know about me: ROAST ME. Don’t hold back. I don’t want superficial jokes; I want you to use what you’ve picked up from the way I write, express myself, and what I reveal here to give me a brutal roast that’s actually grounded in the truth.
1
151
This is how the margin looks for one of my customers in a side project... 💸
If you use @polar_sh you know the revenue from customers. But you can track costs too! So getting nice stats like business margins, top customers by spending, etc. becomes really easy and handy
Conclusion - keep track of your AI costs.
damn I never thought about the possibility that a lot of AI startups only have good margins until people REALLY start using them
as models get better, users burn more token evaporating your margins
so you can only limit them hardcore or switch to open weight models
4
275
In the end LLMs will probably do only of the two things:
1. Take our jobs
2. Or kill us
No better work-life balance option in there 🥲
1
141
That’s my perception also 😅
I think we now have a tooling problem. Agents got capable but still no easy app exists to book a flight, a restaurant, protectively remind me of something.
Astra is basically AGI. And it is absolutely shocking to me how bottlenecked my life still is. By me. All of the things that are stopping me from moving forward are still only resolvable by me. I have more work than ever
1
244
😢 @cursor_ai generates the PR ready for review before I even get to review it. This is really pissing me and my colleagues.
I’ve updated AGENTS.md and user preferences but it’s still not working
3
8
2,282
After nearly 10 years, I cancelled my Todoist subscription this weekend 🫂
Todoist is an excellent app. But I cancelled as I can now manage all my todos and notes inside my personal agent. Having all context is key.
158
This looks really cool 👀
Today we're launching Desert Ant Labs: a European frontier AI lab building on-device intelligence.
18 models across audio, vision, and text. SDKs for Swift, Kotlin, and JavaScript.
No tokens. No logins. Nothing leaves the device.
desertant.com
166
2/3 of today’s top stories are @polar_sh customers 🚀
Great customers trust a great payment platform 😉
1
16
560
This is huge if true. Now the P vs NP
We’re sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
1
170
Shit gets everywhere.
Once a year I tear down the dishwasher and fully clean it, including rails, accessories, and everything
1
2
218
For about two months I have been working on a new personal agent. I want something that is private and has the context from todo, notes, mail, etc.
today it has surprised me because i asked to organize some notes it did it on the first try! 🥲
Shitty feature but it worked
5
165
I remember when antrophic was the good guys… that’s not the case anymore
The whole “20x” in Claude’s pricing is so misleading.
The $100 plan says 5x Pro limits, while the $200 plan says 20x Pro limits. That makes it sound like you’re getting 4x the usage.
But the 20x applies only to the 5-hour usage window. The weekly limit is basically just 2x the $100 plan.
I spent way too much time digging through the internet and Reddit to figure this out. They could’ve just stated the limits directly instead of marketing it as “20x.”
That alone would’ve saved me hours of confusion.
17
7
305
8,996