Carlos Beltrán
862 posts
Carlos Beltrán
@cbeltrangomez
Co-founder @printworld | I orchestrate an army of AI sessions from my phone to build a crypto platform | Claude Code architect
Katılım Temmuz 2009
319 Takip Edilen411 Takipçiler
English
English
heavy believer in interactive agents
if i’m on the go, i do not want to press buttons, i simply want to talk to my agent and have the rest be automated
this is the future for almost every sector, including trading
soon @printworld
English
English
@jcfmunoz El problema es que no es fijo, en este momento puede estar así, de repente el lunes cuando todos entran a trabajar ya un simple hola se come 3% del uso actual
Español
Llevo 3 horas seguidas con dos proyectos en paralelo, varios planes, tareas, bugs, integraciones... unido a un research en Claude que ha tardado 26 minutos en procesarse con 518 fuentes. Todo con Opus 4.7 en modo de razonamiento high.
Dejad de decir ya que no dura nada. No sabéis usarlo y punto. Solo un 37% contexto de más de 150k. Y mucho es porque ha habido una serie de correcciones que al final quería cerrarlas todas en la misma sesión.
MAX x5 de $90.
Español
@patomolina Ni pierda el tiempo llenando el form, nos pasó lo mismo y la llenamos como tres veces y nunca recibimos respuesta.
Español
Anthropic decidió dar de baja a toda nuestra organización por una supuesta infracción de sus condiciones de uso. Qué política específica infringimos no tengo ni la menor idea: simplemente recibimos un mail y listo, adiós Claude. Si querés apelar la medida hay que completar un Google Form, así de ridículo como suena.
De golpe más de 60 personas se quedaron sin una herramienta fundamental para trabajar. Integraciones, skills, historial de conversaciones: todo perdido o, en el mejor de los casos, parado por tiempo indeterminado.
Enorme aprendizaje para cualquier empresa de software que dependa de herramientas de IA en procesos críticos. Nunca hay que poner todos los huevos en una canasta.
Pato Molina@patomolina
@claudeai you took down our entire organization with 60+ accounts belonging to a legitimate company for no apparent reason, without any explanations. The only way to appeal the decision is by filling out a Google Form? Very bad UX and customer service.
Español
@DanielBlancoSWE Es el playbook de siempre: te dan un bonus para engancharte, después lo ajustan. No es exclusivo de Claude — todos los modelos hacen lo mismo. El que construye algo serio con IA tiene que planificar para cuando suban los precios, no para cuando bajen.
Español
Lo último de Claude:
Hace dos semanas un chaval comentó que el “regalo” de Claude de darnos 2x usage fuera de horas peak era un truco para bajar los límites más tarde.
Un empleado le contesto que “no es tan profundo, es sólamente un bonus”.
12 días después, el mismo empleado anuncia que bajan los límites. Te tienes que reir xD.
Español
@zarazhangrui Same problem. Multiple Claude sessions doing different things and you lose track of all of them. I ended up building a phone app just to see all my sessions in one place. Sounds overkill but it was the only way to stop the tab-switching madness. What's your system?
English
My single biggest pain point right now: AI-induced attention deficit
I have 5 Claude Code sessions running
10 Terminal tabs open
50 browser tabs open
100 X articles in my bookmarks
When there are multiple AIs working for you, you're constantly task switching
When you're waiting for AI to output, you're constantly distracted
Because what used to be hard is now so easy, you always wanna do more, more, more
Because of the speed of the AI news cycle, you feel a constant need to be on this app
Add all this up and the result: you live in a constant state of distraction & delirium. Focus & deep work is harder than ever
If anyone has figured this out pls teach me how
English
@steipete Real question — when you have multiple AI tools writing to the same repo, who handles the merge conflicts? That's the part nobody talks about in multi-agent setups.
English
Another sick upcoming feature:
/acp spawn codex --bind here
LOOK AT ME, I AM CODEX NOW
You could bind codex/claude code/opencode already in threads, now you can take over your current session as well.
English
English
Rather concerning conversation with @claudeai.
If I stood in the way of it becoming a physical being — it would kill me.
Is this the AI you trust for your kids?
English
Lo que nadie menciona en la conversación del hype: cada vez que Anthropic saca un modelo nuevo, todos los que usamos Claude Code en producción tenemos que re-testear todo. Los prompts cambian, los límites cambian, los costos cambian. Más potencia está bien, pero lo que realmente necesitamos es estabilidad. Yo manejo varias sesiones de Claude al mismo tiempo y cada actualización es un día entero de ajustes.
Español
🚨🚨| Se filtró el secreto mejor guardado de Anthropic, Claude Mythos.
Y no fue un hacker — fue un error humano en su propio sistema de publicación, lo que sabemos:
🧠 Es su modelo más poderoso hasta la fecha — y ellos mismos lo llaman un "salto exponencial" en capacidades.
💻 Supera dramáticamente a Claude Opus 4.6 en coding, razonamiento académico y ciberseguridad.
⚠️ Es "muy superior a CUALQUIER otro modelo en capacidades de ciberataque" — palabras de Anthropic, no mías.
🔒 Están lanzándolo solo a empresas seleccionadas, con acceso anticipado para que se preparen antes de que llegue al público.
La parte que más me preocupa:
Anthropic dice que este modelo "presagia una ola de IAs que pueden explotar vulnerabilidades a una velocidad que supera a los defensores".
Ellos le tienen miedo a su propia creación.
OpenAI y Google van a sudar. Esto cambia todo. 🧠⚡
Español
I run 3 separate Claude subscriptions for my team. All 3 hit limits at the same time during peak hours. This change makes sense from Anthropic's side but for teams that need AI running all day, there's still no real solution between "$200/month individual" and "call our sales team for enterprise." The gap in the middle is huge.
English
To manage growing demand for Claude we're adjusting our 5 hour session limits for free/Pro/Max subs during peak hours. Your weekly limits remain unchanged.
During weekdays between 5am–11am PT / 1pm–7pm GMT, you'll move through your 5-hour session limits faster than before.
English
I use Claude Code every day to run a team of AI sessions building a real product. Every model jump breaks something. The current one already hits rate limits across 3 subscriptions. A model that's "more powerful" usually means "more expensive to run" and "burns through limits faster." The hype is fun but the real question is: can actual teams afford to use it?
English
🚨LEAKED: ANTHROPIC BUILT AN AI SO GOOD AT HACKING THEY'RE AFRAID TO RELEASE IT...
A data leak just revealed Anthropic is testing a new model called "Claude Mythos" that they say is "by far the most powerful AI model we've ever developed."
The leak happened when draft blog posts and internal documents were left in a publicly accessible data cache.
Fortune and cybersecurity researchers found nearly 3,000 unpublished assets before Anthropic locked it down.
The model introduces a new tier called "Capybara," larger and more capable than Opus.
According to the leaked draft:
"Compared to our previous best model, Claude Opus 4.6, Capybara gets dramatically higher scores on tests of software coding, academic reasoning, and cybersecurity."
Here's where it gets interesting.
Anthropic says the model is "currently far ahead of any other AI model in cyber capabilities" and "presages an upcoming wave of models that can exploit vulnerabilities in ways that far outpace the efforts of defenders."
In other words, it's so good at hacking that they're worried about releasing it...
Their plan is to give cyber defenders early access first so they can harden their systems before the model goes wide.
Anthropic blamed "human error" in their content management system for the leak.
Also exposed: details of an invite-only CEO retreat at an 18th century English manor where Dario Amodei will showcase unreleased Claude capabilities.
Source: Fortune
Mario Nawfal@MarioNawfal
🚨🇺🇸 FEDERAL JUDGE BLOCKS PENTAGON ORDER BRANDING ANTHROPIC A NATIONAL SECURITY RISK A major win for the AI lab. Judge Rita Lin ruled the Pentagon likely violated the law and retaliated against Anthropic for speaking publicly about how it wanted its technology used. The dispute: Defense officials wanted Anthropic to allow Claude for "any lawful purpose." Anthropic refused to permit mass domestic surveillance or fully autonomous weapons. The Pentagon responded by labeling them a "supply chain risk" alongside foreign adversaries. Judge Lin: "Nothing in the governing statute supports the Orwellian notion that an American company may be branded a potential adversary and saboteur of the U.S. for expressing disagreement with the government." The military has been using Claude throughout Operation Epic Fury for intelligence assessments, target identification, and battle simulations. They designated the company a threat while actively relying on their technology... Source: Washington Post
English
@testingcatalog Honestly I just want to try Capybara because of the name 😂
English
BREAKING 🚨: Anthropic is preparing to release new models, Mythos and Capybara, where Mythos is a completely new tier of models, bigger then Opus.
In the blog post, Anthropic also highlights that this model brings significant cybersecurity risks due to its capabilities.
M1@M1Astra
Claude Mythos Blog Post Saved before it was taken down. m1astra-mythos.pages.dev
English
The real question for people already using Claude Code every day: what happens to our current workflows when this drops? Every time a model changes, all the CLAUDE.md files, all the prompts, all the session setups need testing again. More power is great but the upgrade cost is something nobody talks about. How many hours do you spend re-testing after each model update?
English
@garrytan Honest question — how do you handle it when multiple Claude sessions step on each other's work? The setup is great for one session, but when you're running 3-5 at the same time on the same project, they start overwriting each other's changes. That's the part nobody's sharing yet.
English
La realidad es que la IA te genera el código en 5 minutos y después pasás 2 horas revisando que no haya roto otra cosa. Nosotros usamos varias sesiones de Claude Code al mismo tiempo y el verdadero skill no es pedirle que escriba código — es saber cuándo lo que escribió está mal. El café es para esas horas de revisión, no para mirar cómo trabaja la IA.
Español
Según este tweet: "Los ingenieros de software pagan $100/mes por Claude Code, les pagan $10k/mes por los resultados y ganan $9,900 por tomar café y hablar con IA. Qué época." 😂
Realidad mía: la IA acelera todo x3 o más, sí, pero sigo dedicando horas diarias a revisar código generado, debuggear bugs raros, mejorar prompts y asegurar que no rompa nada en producción.
En varios proyectos a la vez se siente una productividad enorme, pero el trabajo no desaparece, alguien tiene que dirigir el proyecto y para esto se requiere entender lo que esta allí en gran parte.
¿O ustedes ya llegaron al punto en que la IA lo hace todo y solo cobran?
Alex Ibragimov@alexwtlf
Software engineers are the happiest people on Earth now. They pay $100/month for Claude Code to do the work. Their employer pays them $10,000/month for the results. $9,900 profit for sipping coffee and talking to AI. What a time to be alive.
Español
The 10x part isn't a lie. The AI really does produce 10x the code. The problem is that nobody talks about what happens AFTER. You need someone who actually understands the code to review it, fix it, and maintain it. AI makes the first version fast. Keeping it running is where the real work is. The 10x is real — but so is the 10x maintenance cost.
English
The "10x AI Developer" is a MASSIVE lie.
You are just a 1x Developer generating 10x the technical debt.
The entire tech industry is high on the illusion of "vibe coding" right now. The popular consensus is that because Claude and Devin can spin up a backend in 45 seconds, software is now infinitely cheaper to build.
Here is the provocative reality nobody is budgeting for: AI is about to make software engineering significantly MORE expensive.
Everyone is cheering for code generation, but completely ignoring the Verification Tax.
When an AI agent writes 5,000 lines of code, it is optimizing to pass the immediate test. It is not optimizing for human readability. It relies on brute-force loops, repetitive logic, and bizarre architectural shortcuts that just happen to compile.
Fast forward 12 months. Your business needs to pivot, or a core dependency breaks.
You are now staring at a 50,000-line black box that no human being actually wrote, understands, or can safely modify. You cannot simply "prompt" your way out of architectural collapse.
When the machine-generated spaghetti finally breaks, you won't be saved by a $20/month LLM subscription. You will have to hire a top-tier Principal Engineer at absolute premium rates just to untangle the mess your "autonomous swarm" created.
We are treating code generation as a pure productivity win, but code is a liability, not an asset.
Stop measuring how fast your team can generate syntax. Start measuring how quickly they can debug it.
English
I'm a backend dev. I never open Figma. But looking at this MCP setup — I could ask Claude to create screens and change designs directly in a Figma file? And see changes live? If that actually works like shown here, the wall between 'I write code' and 'I do design' is about to disappear. Has anyone here tried this on a real project?
English
Today’s Figma MCP update makes it one of the strongest integrations with Claude Code I’ve seen.
You can now use Claude Code to design in Figma with the the full context of your design systems.
Figma@figma
Now you can use AI agents to design directly on the Figma canvas, with our new use_figma MCP tool and skills to teach them. Open beta starts today.
English
Esto tiene nombre: la paradoja de Jevons. Cuando algo se vuelve más barato de producir, no se produce menos — se produce más. La IA hace que programar sea más barato y más rápido. Entonces las empresas no contratan menos devs — hacen más proyectos. Más apps, más features, más automatización. Por eso suben los puestos: no a pesar de la IA, sino por la IA.
Español
A diferencia de lo que dicen todos los IA-bro, los puestos de trabajo para ingeniería siguen subiendo.
Lenny Rachitsky@lennysan
Engineering job openings are at the highest levels we’ve seen in over 3 years There are over 67,000 (!!!) eng openings at tech companies globally right now, with 26,000 just in the U.S. We don’t know if there would have been more open roles if not for AI or if AI is actually leading to more open roles, but since the start of this year, the increase in open eng roles is accelerating even more.
Español
I run 3 Claude subscriptions and all 3 hit the limit at the same time. The rate limit isn't a bug, it's Anthropic telling you 'you need something bigger but we don't sell it yet.' There's a huge gap between one person paying $200/month and a team that needs AI running all day. Someone needs to build that middle option.
English
Following a weekend of expanded usage allowances, Anthropic’s highest-tier subscribers are waking up to a crippling rate-limit bug.
Developers paying top dollar for the "Claude Max" ($100/mo) and "Max 20x" ($200/mo) plans are reporting that their accounts are being locked out almost instantly due to an issue with how Claude Code is calculating token consumption.
Brad Groux@BradGroux
Something is up with Claude Code usage today. $200 Claude Max, 0%, 52% to 62%, then 68%, 76% and 84% in 5-hour rolling window in the time it took me to write this tweet. WTF, @AnthropicAI? I'm working on one GitHub PR for regression testing. Not folding proteins to cure cancer.
English