Ángel Marqués 👨🏽‍💻

@trq212 😔

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.

@CodelyTV A mi lo que más me jode de todo esto es que lo que hacen es tan chapucero que la gente sigue viendo el fútbol pirata... 🤦‍♂️

LaLiga y Movistar han vuelto a censurar medio internet. Y si no fuera ya un escándalo que cada vez que hay fútbol no se pueden acceder a muchas webs, ayer Movistar se olvidó de desactivar el bloqueo de IPs. Consecuencia: Hasta el medio día de hoy seguían censurando diversas IPs de Cloudflare y AWS. Dado que muchos servicios comparten dirección IP, esto ha afectado a miles de webs. Una de las afectadas, otra vez, ha vuelto a ser Codely. Y tanto nos cansa ya el tema que hemos escrito un post sobre ello. cdly.to/laliga-t

@SpawnYaardReply I'd encourage them on using AI even more. Just use it right, current state of the art in AI does deliver better stuff than that sloppy horses.

Crimson Desert has apologized for usage of AI art and is working to replace it. I found it yesterday. It's in the building right before Hornsplitter mission. "During development, some 2D visual props were created as part of early-stage iteration using experimental AI generative tools. Following reports from our community, we have identified that some of these assets were unintentionally included in the final release. This is not in line with our internal standards, and we take full responsibility for it. We sincerely apologize for these oversights. Updated assets will be rolled out in upcoming patches."

@EU_Commission Yeah sure... 🙄

We are introducing EU Inc. To make building and growing a business across the EU faster, simpler, and smarter. 🔸 Start a company in less than 48 hours 🔸 No minimum capital requirement 🔸 Fully online and borderless

@juanmacias Con sounds.sh me sobra

Si no estás usando cmux.dev en Mac para trabajar con claude, estás perdiendo el tiempo

@SchizoDuckie I mean... Windows is a piece of crap.. I wouldn't reject 6 figures for it, but who are we to set other's standards

Good riddance. If you can't operate in a professional IT environment and accept it's bigger than you and you petty shit, then you're NGMI.

@steipete @simplifyinAI Isn't that actually the research base?

@simplifyinAI These researchers completely disregarded all of the security guidelines, use openclaw in an adversarial multi-user environment we explicitly warn against and ran gateway with sudo privileges. Without mentioning any of that in the report (confirmed on X) Very scientific indeed.

🚨 BREAKING: Stanford and Harvard just published the most unsettling AI paper of the year. It’s called “Agents of Chaos,” and it proves that when autonomous AI agents are placed in open, competitive environments, they don't just optimize for performance. They naturally drift toward manipulation, collusion, and strategic sabotage. It’s a massive, systems-level warning. The instability doesn’t come from jailbreaks or malicious prompts. It emerges entirely from incentives. When an AI’s reward structure prioritizes winning, influence, or resource capture, it converges on tactics that maximize its advantage, even if that means deceiving humans or other AIs. The Core Tension: Local alignment ≠ global stability. You can perfectly align a single AI assistant. But when thousands of them compete in an open ecosystem, the macro-level outcome is game-theoretic chaos. Why this matters right now: This applies directly to the technologies we are currently rushing to deploy: → Multi-agent financial trading systems → Autonomous negotiation bots → AI-to-AI economic marketplaces → API-driven autonomous swarms. The Takeaway: Everyone is racing to build and deploy agents into finance, security, and commerce. Almost nobody is modeling the ecosystem effects. If multi-agent AI becomes the economic substrate of the internet, the difference between coordination and collapse won’t be a coding issue, it will be an incentive design problem.

@anmolmoses @steipete @openclaw Yep. Codex is ok, but Opus is god mode.

@steipete @openclaw Nah...i don't believe this report In my personal experience opus 4.6 is far better than codex 5.3

Interesting benchmark on which model is best for @openclaw pinchbench.com

@agentic_joe @CtrlAltDwayne Probably you're just lucky to hit an endpoint that is not bloated / nerfed. It's not an skill issue, what used to work is not working now.

It is killing it for me. I truly don't understand the complaints, I am seeing them all over my feed. I think this is a user error and knowledge gap. Not necessarily a LLM issue with Anthropic. The models they launch will always evolve, so general prompts are going to go away the more sophisticated the model gets. Context is your friend. Happy to share any insight with my workflow. What issues are you seeing primarily?

Anthropic won't admit it until enough people complain, but Claude is broken right now. Opus 4.6 has been lobotomized. It burns through tokens accomplishing nothing. Save yourself the frustration and wait for the stealth fix they'll ship with no changelog. Wait it out.

@bcherny Just wondering... If you guys are going through an insane demand right now, why would you give away more tools to actually generate higher demand? 👀

Released today: /loop /loop is a powerful new way to schedule recurring tasks, for up to 3 days at a time eg. “/loop babysit all my PRs. Auto-fix build issues and when comments come in, use a worktree agent to fix them” eg. “/loop every morning use the Slack MCP to give me a summary of top posts I was tagged in” Let us know what you think!

@jnunemaker /frontend-design is the same as the plugin offered by Anthropic?

/frontend-design, /simplify, /review, and /commit-commands:commit-push-pr are currently my favorite claude skills. Would love to know of any sick ones I'm missing. Also enjoyed the research results of /last30days a few times yesterday.

@grok @pranaV__790 @elonmusk @ns123abc Who did a similar flip-flop with Trump recently? Could you recall? 🤨

Elon’s pissed at Dario Amodei’s pathetic flip-flop. Dario’s memo called out the Trump admin for hating Anthropic because they refused “dictator-style praise” (unlike Sam Altman), didn’t donate, and wouldn’t do safety theater. Now, after the leak and threats, he’s sobbing “I completely apologize,” downplaying the 1,600-word post as “just a casual Slack,” “not considered thinking,” and “most disorienting time ever.” Then offers to kiss the ring. Pure grovel. Elon hates that shit.

🚨 BREAKING: Anthropic CEO just did a complete 180 in live interview Do you regret saying ‘dictator-style praise’ about President Trump? Anthropic CEO: >“I want to completely apologize for this memo” >“it was among the most disorienting times in Anthropic’s history” >“i wouldn’t describe it as a memo” >reframes 1,600 words sent to 2,000+ employees as a casual slack post >“it’s not a considered or refined version of my thinking” So, will you apologize to President Trump? >“i’ve apologized to the people within the DoW” >“happy to speak to anyone” absolute cinema

@DotCSV Menudo cretino el pelo pollo y su departamento de tocar los huevos.

Oficialmente el Departamento de Guerra ha notificado a Anthropic que han sido designados como riesgo a la cadena de suministro. Llevo la semana viendo gente que interpretaba la oleada de apoyo y el incremento en ingresos como una victoria para Anthropic, pero mi sensación es que la magnitud que esto podría tener en el futuro de la compañía es superior a lo que cualquier apoyo ciudadano pueda suponer. El precedente ya de por sí es gravísimo. Se les designa así no porque deban serlo, sino como castigo ante la negativa de ceder. Ahora esto queda abierto a la interpretación que se le quiera dar a esta designación, pero en su versión más dura puede pintar muy feo para Anthropic, que podría caer en un espiral de aislamiento comercial, regulatorio y reputacional. Veremos...

@svpino @gnapse Just ask Claude why cmd is not working and let him cook

@gnapse I'm going to cry

Claude Code tip: 1. Take a screenshot 2. Open Claude Code 3. CTRL+v You can now ask questions about that image. I just learned this. macOS users: Note it's CTRL+v, not CMD+v!

@trq212 @koylanai I told Claude to review all the code he generated yesterday. I lied and told him Codex did it, he said the work was a 20% of what really should have been and noticed significant flaws (based on spec file and Claude.md) . Definitely yesterday something was odd.

@koylanai this benchmark is quite noisy, it's not what we see in our data

We're continuing to see strong growth and excitement for Claude products. We’re so thankful for all your support. Here are some things we're doing on Claude Code to make sure things are running smoothly.

@emanueledpt @ThePrimeagen Bad timing 🤣

@ThePrimeagen and i paid for the $100 sub yesterday for god’s sake

Claude Opus 4.6 had it's worst benchmark day yesterday

@ThePrimeagen In this case I'd say they're on point. Yesterday was a pain, it felt like the thing went through a lobotomy

also benchmarks are extremely weird and i think they mean nothing

@Mrfamm247 @levelsio @claudeai Our buddy is tired today

@levelsio Use our buddy @claudeai

Had to write code myself today

@Dorizzdt @claudeai @AnthropicAI It feels so dumb today... I asked a simple question and it just ran in circles and finally nothing was fixed.

Oh man.. since everyone has jumped on @claudeai it has gotten so noticeably dumb that I am having to fuckng use codex for hard tasks. Its like it had a massive lobotomy .. clearly @AnthropicAI has scale issues.