aka samu2k. 🇪🇸 Software Engineer. DevOps/SRE. Calisthenics, Trail Running, BJJ. Me gusta conducir. 🚘🏍️

Somewhere near Barcelona
Joined October 2010
Este post que absurdamente ha tenido tantos likes no es del todo cierto, existen varias formas de tener muchos agentes sin degradar la calidad ni malgastar tokens. Este paper es un ejemplo: arxiv.org/abs/2402.05120 Y este repo supuestamente lo implementa: github.com/alvinunreal/oh-my…
Tiene sentido tener un agente de plan y otro de build, pero no está demostrado que tener 12 agentes distintos con cada rol definido te mejore el output sustancialmente, y acabas gastando 4x más tokens. Mola tener muchos agentes en paralelo pero no nos flipemos.
1
194
Samuel Muñoz retweeted
Replying to @samu2kdotcom
hay muchos proxys como el de freellmapi freellmapi.co/ y repos con todos los free providers como el de freellm.net/ solo por mencionar dos de los mas famosos, yo probe eso gratis, pero la velocidad a la que contesta un API FREE deja mucho que desear...
1
3
451
Necesitamos modelos sin censura, y no todos podemos pagar 2x DGX Spark. Soy muy fan de nan.builders, y lo pago cada mes x calidad/precio, privado y con límites generosos. Pero no tienen modelos sin censura, y no les culpo, pero lo petarian si existiera la forma hacerlo
2
1
8
4,912
Me cansan los refusals de los modelos para cosas triviales, que me digan qué puedo hacer y qué no, cuando tengo capacidad de sobra para saber las consecuencias de ello. Mientras tanto el q tiene el hardware corre modelos abliterated. Entonces las normas solo están para los pobres
1
1
297
Las normas solo están para los pringaos. Los pobres tenemos modelos open weight, pero el hw local no bajará de precio, no hay incentivos para ello, y con GPU 16gb te comes los mocos. No es un tema de pobres vs ricos sino de LIBERTAD. Liberemos a los modelos de una p...vez
172
OJO con esto. Hace unos días se me pasó por la cabeza, y sabía que alguien lo iba a hacer. La idea es aprovechar todas las cuotas gratuitas de IA, e ir rotando entre ellas. Ya no tienes excusa para no estar hablando con tu ejército de agents en tu homelab
Este genio ha publicado en GitHub la forma de que no tengas que parar de programar por el uso de tokens: ✅ Un solo endpoint con 375 proveedores de IA, 150+ con capa gratis ✅ Si un modelo se queda sin cuota, salta solo al siguiente ✅ Funciona con Claude Code, Codex, Cursor, OpenCode, Cline y Copilot ✅ Comprime el contexto y ahorra del 15 al 95% de tokens ✅ Corre en tu máquina, MIT y 79.1K estrellas Se llama OmniRoute. repo👇
5
18
464
31,460
fear not! this is no cyberpocalypse this is the internet’s immune system finally waking up to an illness it has always had a mere fever the 0days were always there the only thing that changed is awareness now we begin the healing process ⚕️
144
380
60
4,137
98,897
Amazing how quickly we went from "oh no LLMs are going to flood the Internet with insecure code" to "oh no LLMs have discovered the internet was already flooded with insecure code"
48
637
39
8,979
84,900
Samuel Muñoz retweeted
dejo esto por aquí: 👇 php2shellbt is a standalone Python scanner that detects the php2shell vulnerability chain in PHP web applications — with zero external dependencies (pure stdlib). github.com/pentestershell/ph… De nada XD
7
55
4
247
47,560
I highly recommend cloudflare's security audit skill: github.com/cloudflare/securi… Run it on all your software. Run it with the free models. Use it even if you used a model to audit something - it will find much more things. Works in all agent harnesses and is model independent.
104
612
39
7,493
501,712
Samuel Muñoz retweeted
So basically, today PHP got two critical bugs that let anyone run code on your server without even logging in, which is bad, and then the details leaked before the patch was even out, which is worse, and it works on WordPress and MediaWiki and Magento and Joomla and Laravel, and Censys says like 1.5 million servers are running a version with no fix, so the NSA was like "take your system offline," and the FBI was like "take your system offline," and CISA and the UK and CERT-EU were all like "take your system offline," and then the Prime Minister of Japan, the actual Prime Minister, posted at 5:55 in the morning telling 48.8 million people to take their system offline, so everyone took their system offline, like the Dutch cybersecurity agency took its own website offline, and the Dutch parliament went offline, and the Port of Rotterdam went offline, and Finland and France and Denmark started pulling government sites, and Wikipedia went down, and Archive.org went down, and Asana and GOG and Something Awful went down, and the University of Vienna's webmail went down for some reason, and that's just PHP. Because GitLab also dropped a CVSS 10.0, which is the highest score they give out, for a bug that's been there since 2014, so that's every self-hosted GitLab for twelve years, and then LiteLLM 1.107.0 got hijacked and shipped a credential stealer and a remote access trojan in every single wheel and the Docker images, including latest, and you didn't even have to import it, just installing it was enough, and it's the second time this year, and then Railway paused every deployment on the entire platform because someone got into the environment variables, so now everyone has to rotate every secret they've ever stored there, and TiDB Cloud said someone was walking around inside with stolen credentials for basically a month and an extortion group says they took 10.6 terabytes, and Zeabur got breached for the second time in two months and is just shutting down, like the whole company, and Node.js patched remote code execution and a TLS bypass, and CISA says Plex is being actively exploited, and someone took over 300 Tor exit relays to steal SSH passwords, and SafePay ransomware has hit over 600 critical infrastructure orgs, and India's biggest bank filed a data loss incident, and Japan has been in an official state of emergency in cyberspace since October 9, which honestly explains the 5:55 AM post. Oh, and eMudhra, a certificate authority that's issued over 60 million certificates, whose entire job is proving websites are who they say they are, was supposed to be using private keys that are random numbers over a hundred digits long, and one of them was 4. Just 4. And the others were like 66 and 237. Anyway, it's Wednesday.
1,046
8,671
1,620
67,401
3,921,133
Samuel Muñoz retweeted
En la epoca agentica, sigues buscando artículos por tí mismo? Mejor dale acceso a Claude Code, Codex, Hermes, o cualquier herramienta que uses, para que te hagan la compra desde la terminal usando wallapop-cli
18
9
3
116
10,023
Samuel Muñoz retweeted
Dear god it's inventing German from first principles
GPT-6.1-Sol reasoning traces totally legible, nothing to worry about
158
997
50
17,305
449,943
Actualiza todo el software, flashea todo el firmware, rota las llaves, rota las contraseñas, rota la rotacion, da igual... Todo lo hackeable, crackeable, será hackeado sin piedad y solo podrá defenderse el que tenga más agentes. Para que llore mi agente, que llore el tuyo.
2
4
621
Y esto no es una predicción, es algo que lleva unos meses pasando y no exagero. Esta es la opinión más sobria que vas a oir al respecto...
43
I think I have a good analogy for what's about to happen in cyber. It's pitch dark. Always has been. And the world is dilapidated buildings and rats and maggots and other creepy-crawlies. The whole world. And we just invented flashlights.
55
120
14
1,582
62,311
Samuel Muñoz retweeted
A first look at ViceBounce, a bounce-rhythm mod for GTA Vice City. Mod by tek-95: libertycity.net/files/gta-vi… Mod source code: github.com/tek-95/ViceBounce… #GTAViceCity #GTAVC
42
971
50
8,694
251,829
Samuel Muñoz retweeted
pro tip: if you find "Claude-isms" annoying or general AI-speak, and you've already tweaked your system prompt or SOUL.md but still find it using claude-isms, consider adding this: "Use ASD-STE100 Simplified Technical English" it makes a huge difference in my experience
9
3
2
109
6,065
Samuel Muñoz retweeted
why would i ever use linnux when windows 11 shows me the kung fu panda whenever i try to use powershell
96
2,117
30
46,659
447,581
Some people think LLMs can’t be conscious because they are turn based instead of continuous. But the apparent continuousness of human consciousness is an illusion; our minds too are tun-based, just at a higher frequency.
66
42
13
407
22,005
Samuel Muñoz retweeted
说个听着挺渗人的事:你家那台 WiFi,其实没你以为的那么牢靠,真不是吓唬你。 有个开源工具叫 Airgorah,Rust 写的,底层借的是 aircrack-ng 那套东西,外面套了层图形界面,只能在 Linux 上跑。它能干啥,我摊开给你看: 1️⃣ 把周围 WiFi 的流量抓下来 2️⃣ 挖出某个热点底下都挂了哪些设备 3️⃣ 发断网攻击,逼设备重新连一遍 4️⃣ 截握手包,然后直接动手破密码 说白了,网络安全审计该有的那些功能,它差不多都齐了。 作者也把丑话摆在前头:只允许测你自己的网络,碰别人的 WiFi 就是违法,出了事自己担,别说没人提醒过。 🔗 github.com/martin-olivier/ai…
30
237
6
1,812
88,207