
SSWE @HubSpot. Proud dad. Love 🏀, MTG 🧞♂️and solving problems🧑💻one prompt at a time.
Liege, Belgium
Joined April 2010
- Tweets3.8K
- Following506
- Followers295
- Likes11.8K
Javi Carretero retweeted
human-in-the-loop
└ human-on-the-loop
└ human-near-the-loop
└ human-cc'd-on-the-loop
└ human-legally-liable-for-the-loop
Day 28/ tibo's game has landed
i spent $5.99 on this domain, but i had a blast making it with grok bot and cursor! i hope you enjoy it and find all the easter eggs 🤍
if i did it right you should be able to play it inside of X too
tibo.dog
Javi Carretero retweeted
A CVE (Common Vulnerabilities and Exposures) is a publicly disclosed security flaw.
They increased following a slight upward slope all the way to early 2024, when there was a sudden small increase... then starting in 2026, we went full exponential.
In most instances, the vulnerabilities were pre-existing, so much of the increase is just our better ability to find the faults.
This is of historical importance. We have never witnessed anything of the sort.
Javi Carretero retweeted
Un experto con IA siempre será capaz de ofrecer mejores soluciones que un no experto con IA.
Por eso creo que aprender es cada vez más importante, debemos exprimir la IA todo lo que podamos pero eso no nos tiene que robar al interés y la curiosidad de aprender por nosotros mismos.
Javi Carretero retweeted
Calling it now.
AI is going to usher in a whole new era of software development. I’m calling it Software as a Service.
People will want to use these tools but they’ll realize they don’t want to spend the time to maintain and update them over time.
This will create a whole new category of businesses and entrepreneurs that can take on that task for businesses that would rather work on their core products.
AI is changing everything. This is a whole new paradigm of thinking about the way we interact with software.
Javi Carretero retweeted
Tras la rueda de prensa del Ministro, la cuestión no será quién se equivocó esta mañana. Hoy toca el respeto y el silencio por las familias. Pero mañana tocará preguntar.
Veamos primero LOS HECHOS
1. El sistema de señalización de la línea estaba averiado. Al caer, entró un modo degradado que obliga a circular a marcha a la vista. Según las primeras informaciones, hasta ocho trenes quedaron detenidos en la zona.
2. Un maquinista se encontró una señal en rojo, pidió autorización al centro de control de Atocha para rebasarla, y delante había un tren parado.
3. El Gobierno apunta a error humano de un técnico de Adif. Puede que lo sea. Pero el error humano es el final de la cadena, no el principio.
4. El sistema que falló es el LZB, instalado en 1992, cuando se inauguró el AVE para la Expo. Madrid-Sevilla era la única línea de alta velocidad española que seguía con esa tecnología. Todas las demás ya operan con el estándar europeo ERTMS.
5. Su sustitución se licitó en 2020. La licitación tuvo que repetirse. Se adjudicó en septiembre de 2021, con 52 meses de plazo. En abril de 2025 el Ministerio fijaba el objetivo: la línea cumpliría su 34 aniversario completamente renovada y con ERTMS 2. Es decir, debía estar terminada este año.
6. No lo estaba. En el momento del choque, el ERTMS todavía no era el sistema operativo principal de la línea. Y mientras tanto se siguió explotando a pleno rendimiento, con demanda récord y tres operadores compartiendo una infraestructura concebida para uno.
7. Hay 700 millones invertidos en la renovación integral de esa línea, según el propio Ministerio. Sirvieron para renovar vía, traviesas, desvíos, subestaciones. No sirvieron para completar la señalización.
8. Y aquí está el problema de fondo. España construyó la red de alta velocidad más extensa de Europa mientras su primera línea seguía protegida por un sistema de 1992. La inauguración rinde en titulares. El mantenimiento no se inaugura.
9. Hace nueve meses, 46 fallecidos en Adamuz. La misma línea. La investigación sigue abierta. Dos catástrofes ferroviarias en el mismo corredor en nueve meses, en el país que presumía de tener el mejor tren de Europa.
Y ahora LAS PREGUNTAS:
A. ¿Quién autorizó los sucesivos retrasos del ERTMS y con qué informe de riesgo?
B. ¿Qué análisis se hizo sobre explotar la línea en modo degradado durante las obras?
C. ¿Por qué un protocolo permite sustituir una protección automática por una autorización verbal sin una barrera de respaldo?
D. ¿Qué recomendó la comisión de investigación tras Adamuz, y qué se aplicó?
E. Las recomendaciones de seguridad las emite el investigador, pero quien decide aplicarlas es el investigado. ¿Quién verifica que se cumplen, y qué ocurre si no se cumplen?
Hoy no toca responderlas. Toca acompañar a quien llora. Pero que nadie nos convenza de que esto fue mala suerte, ni de que se resuelve señalando a un operador en una sala de control de Atocha.
Los accidentes no empiezan el día que ocurren. Empiezan años antes, en un expediente que se aplaza, en una licitación que se repite, en una prioridad presupuestaria que se cambia. En unos presupuestos que no se presentan ni se aprueban.
Hoy hemos visto el final de esa cadena. Falta saber dónde empezó.
Conviene ir anotando las preguntas. Dentro de dos semanas muchos querrán que no se hagan.
i own a relatively new (2025) samsung tv.
i kept getting weirdly targeted ads.
i tried reversing the firmware but it was aes encrypted and no keys unless tear it apart and do uart/jtag magic.
zero days it is. root shell on new samsung TVs!
discovered PRIVACY HORROR. write up?
here's how i shipped 2,500 PRs last month to production
this was originally supposed to be for Cursor Compile in London. i couldn't make it since i was livestreaming for Grok @Bot Galaxy so i'm making it available for free here on X! watch it on 2x speed, i talk slowly
i love writing. i never let agents write for me. whether it's docs, or tweets, or shitposts, or even prompts (okay maybe that one's harder as agents manage other agents)
there's a simple reason for that: i still strongly believe that writing is thinking. it's a skill i hope to never give up. it's also important to me that whatever i say is always authentic, mistakes and all. my favorite posts are the ones that i just full send without any editing
i suspect it also helps with prompting. being able to clearly express what i want, and knowing the vocabulary i need to be able to direct my agents does seem to make for more effective usage of agents. in the same way that engineers have an advantage when we prompt agents to write code, because we are intimately familiar with the words that make up the discipline
knowing the right word incantations to get an agent to produce a desired goal almost feels like spellcasting. as models get more capable, and that input box looks more and more like a magic box that can output anything, words become the interface between you and the genie
there may come a time when we give up the keyboard in exchange for brain-computer interfaces. but until then, i'm writing every word.
Javi Carretero retweeted
I’m really trying to pretend this is true so RC doesn’t get steamrolled, but the more I think about it, the less I buy it.
1. Historically, the production of many types of goods have decreased dramatically in complexity and it didn’t transition to everyone just building it themselves. When electric tools were invented, did everyone transition to making their own furniture? Did the word processor end the book industry? Did the camera end the art industry? None of these quite match what we’re seeing but I haven’t found a case where humans have de-specialized.
2. Most consumers are exactly that, consumers. They are much lower agency than your average software person, probably lower creativity, and aren’t going to start building their own novel solutions to problems all of a sudden. Perhaps the models get so good they can do it somewhat transparently, but that would still require dictating requirements to the models, which itself is not something most consumers will be great at. And part of making “good” software is being opinionated. Models, due to the efforts of alignment, are limited in what they can say no to. The user asks for something that makes the software worse? “You’re absolutely right!” I’m not sure that taste and alignment are fully compatible.
3. Nothing ever happens. I think software will evolve, and what was a successful app three years ago will be very different three years from now, but it just doesn’t seem likely that after 10,000 years of increasing specialization and craftsmanship, humans will just resort to self reliance all of sudden. We want the expert opinion of other humans. We are interested in novel tools and experiences. Software products often bring with them community and identity that won’t go away. I just think an amplified version of the current trajectory more plausible than a complete reversal.
Tell me why I’m wrong!
Javi Carretero retweeted
Amazon banning agents is the first opportunity I've seen since Amazon was founded for a startup to create an Amazon competitor. People will want agents to buy stuff for them. It will be one of the main use cases. And they won't want to use some Amazon-supplied agent to do it.
Javi Carretero retweeted
I got some issues with this.
The $132m/year is not real at all. Closer to $3m at today's prices, and as low as $1.2k within a year.
Yes, $132m -> $1.2k. Hear me out.
1. The price listed is not accurate at all.
$11m/month is just wrong. Real "api cost" is prolly $100k-$500k per month range.
The 1.5T tokens number includes cache'd tokens. I looked at my own usage, and 33.5B tokens came out to $0.56/m (under 1/10th of Peter's $8/m estimate).
I'm primarily using expensive models like Opus and Astra, and "1.5t tokens" would still have cost me ~$800k. If I used cheaper models and hit cache more, easily could get as low as $100k.
2. Intelligence is getting cheaper incredibly quickly
I know our LLM bills are going up and everyone is stressed about it. Totally reasonable! It's not because intelligence got more expensive.
Artificial Analysis tracks how "cost at a given intelligence level" drops over time. It used to drop at ~10x/year. Now it's at 30x lower in 5 months. (see image)
Assuming the $100k from before, cost could get as low as $3k/month in a few months, and if trends continue, $100/month by this time next year.
Obviously there will be smarter models and we will move to them, but replicating her experience today with equivalent models will be $1.2k/year by next year.
3. Nobody said this is normal
Lauren is the 0.0001% of "agentic engineer". Measuring her work as "$ in -> software out" is reductive and wrong.
Her spend should be looked at as a combination of product/eng, research, FDE work (pstack), and marketing. Worth $350k/day? No. Worth a mil or two a month? Absolutely.
She is exploring what happens when you ignore the cost of intelligence. She's not doing this because Cursor expects every company to spend $10m/year on tokens.
She's doing it because that $10m in tokens is going to cost $10,000 in a year or two.
poteto's cursor profile is public. She's used 1.5T tokens in the last 30 days.
Assuming an avg rate of $8/M that's $384k/day or $11m. In the last 30 days. Per year thats $132M.
Let's assume Cursor has some insane deal with Anthropic and gets 90% off. That comes out to $38k/day or $1.1M in Sept/Oct.
$13m will get me 26 sassy staff-level all stars for $500k/yr
Readers added context they thought people might want to know
The Cursor profile at cursor.com/@lauren shows 1.5T tokens used. Cost estimates at $8/M do not account for prompt caching common in agent coding, where cache reads are billed at ~10% of input rates per analyses of Anthropic pricing. This results in lower effective costs.
cursor.com/@lauren chudi.dev/blog/claude-pr… carlosaragon.online/blog/anthropic…
Javi Carretero retweeted
Not sure who needs to hear this (well, I guess I am), but „my agent did this“ is never an acceptable answer if a system you built is not working correctly. That bucket stops with you.


























