Nabetse

vibecoders using claude, chat gpt and gemini for the same project be like:

this is the MOST important 4 minutes you’ll watch on AI this year. anthropic built a model so good at finding vulnerabilities they didn’t release it to the public. >CLAUDE MYTHOS PREVIEW it’s unreleased to the public and here’s what it did in a few weeks: >found a 27-year-old vulnerability in OpenBSD >caught a 16-year-old flaw in FFmpeg that automated tools missed after 5 million tests >chained together multiple linux kernel exploits autonomously. no human steering. AWS, google, microsoft, apple, nvidia, crowdstrike, JPMorgan. all got access. Anthropic committed $100M in credits to let these companies hunt vulnerabilities in their own systems before attackers do. >93.9% on SWE-bench verified. >77.8% on SWE-bench pro. nothing else is comes remotely close. Anthropic just pulled away in this AI race…

Metadata del post: Autor otolz_, ID 2030838182556708957, 9 mar 2026 02:48 GMT. 310 likes, 49 reposts, 28k views + video demo. El software es WiFi-DensePose (RuView), 100% open source en github.com/ruvnet/wifi-de…. Basado en paper CMU arXiv:2301.00250. Código: extrae CSI (fase/amplitud de señales WiFi). Sí, exactamente mide ondas con rebotes/multipath en el cuerpo, procesa con filtros + red neuronal/graph transformer para estimar pose en tiempo real. Corre con Docker o ESP32.

Soy muy fan de ese niño.👇🥰😇

Google has paid $250000 for Sandbox escape report and POC is public issues.chromium.org/issues/4125787…

Extraño cuando mi cerebro entendía todo esto. 🤓

@zekibahce Birds are not real

The current top US-based hacker on Hackerone's leader board is "Xbow". You might wonder what the heck that is. No, it's not an experienced hacker. It's not even a human. It's an AI agent, that's fully automated trained on real vulnerability data. The future with AI is scary...

Top 8 MCP servers for cybersecurity: 1. Wiz MCP Server — cloud security 🔗 lnkd.in/gqYNzzRZ 2. Cloudflare MCP Server — Infrastructure and security 🔗 lnkd.in/gRNY5pkZ 3. PortSwigger MCP Server — Webapp security testing 🔗 lnkd.in/gcD83ftM 4. GitHub MCP Server —  Developer security & automation 🔗 lnkd.in/gYF26mVB 5. Akto.io MCP Server — API Security 🔗 akto.io/mcp 6. Check Point Software Point MCP Server — Network and threat prevention 🔗 lnkd.in/gV5CsfgQ 7. Google MCP Security — Cloud security and threat intelligence 🔗 lnkd.in/guTsdHdZ 8. Elastic Security MCP Server — SIEM and detection 🔗 lnkd.in/gQe5aAvb

Ayer falleció George Wendt, al que la mayoría asocia con su personaje en una de las series más míticas de la televisión. Al entrar le gritaban: "NORM" y el respondía con una genialidad. Era el lugar en el que todos saben tu normbre. ¡Brindemos! Empieza el #MakeEmLaugh!🍺 🧵⤵️

@NicolasToro2024 @petrogustavo @GustavoBolivar @CarlosCarrilloA Puro lame suelas en ese video

En Pasto salimos a marchar porque es un derecho de todo ciudadano.

@javilop Wow magnífico

⚡ Magnific en la gran pantalla. ¡POR FIN PUEDO HABLAR DE ESTO! El equipo de VFX de Here (dir. Robert Zemeckis y protagonizada por Robin Wright y Tom Hanks) ha usado Magnific para sus FX 🤯 Para contaros todo (y más), he entrevistado al supervisor de VFX Kevin Baillie 🧵👇

🔴 UNA IA CONSCIENTE Hoy voy hablaros de cómo podríamos QUIZÁS conseguir algo que el ser humano posiblemente jamás deba intentar: Un método para que una IA cobre CONSCIENCIA y adquiera el estatus de una super inteligencia (ASI). Teoría en la que he trabajado varios meses 🧵

AI influencers are quickly starting to become big business now 🇺🇸 American AI influencer @lilmiquela makes $10 million per year 🇩🇪 German AI fashion model @noonoouri makes $500,000 per year 🇿🇦 South African AI supermodel Shudu makes $2,000 per sponsored post Last year the industry was valued at $6 billion and by 2030 it's projected to grow to $45 billion in money spent by people on virtual influencers A LOT of people asked me to make it possible to create their own AI Influencer from scratch So this week I've added this ability to Photo AI You can design your AI influencer based on where they're from, how they look, etc. then once you created it, you can take photos and videos with them and even make them talk This AI influencer below does not exist and was made with Photo AI in 30 seconds for example:

@hipdead010 Todo envío de tokens de autenticación deberia ser a través del payload en solicitudes POST. Pero bajo ninguna circunstancia deben enviarse como parámetros en la URL mediante GET, ya que pueden quedar en cache, ser indexados por crawlers, o quedar en los logs del servidor.

parece que la tendencia DeepSeek tiene malas practicas, en este caso podemos ver donde varias autorizaciones de autenticación con google, se filtra un JWT, el resto ya lo conocen. wb = waybackmachine

@val__lop_ En el puerto es obligatorio que todas las casas y negocios tengan pozos sépticos. ¡Incluso hace un par de años regalaron pozos a toda la comunidad! Pero pocos los instalaron… ¿Será que antes era como una escena de película de terror ambiental? 🌊💩

Increíble el color que tiene el agua que baja desde el puerto hacia la cocha, contaminación totaaal🙃

Saw a guy hunting for threat actors in the network today. No SIEM. No IDS. No EDR. He just sat there. Watching traffic in wireshark. Like a psychopath.

The three areas of need are: 1. Agent Authentication & Authorization 2. Prompt Injection-related attacks 3. Secure Agent Architecture For #1, Agent authentication can likely be put into existing stuff like okta etc. Authorization is the really hard part and will need to be solved for every app. Companies like @browserbase will probably shine here. On #2, Prompt injection is still unsolved. Most AI apps are shipping with very little red teaming (and security testing). Companies like @haizelabs can find those flaw near instantly, but at the very least, people should be having pentesters look into them. The space for #2 to be an issue is HUGE though. The attack surface is the entire internet AND the entire real world since embodied agents are coming. My major hope is that models just get smart enough to not fall for prompt injection/jailbreaks. Humans are gullible but for the most part, prompt injection and jailbreaks are obvious to humans. For #3, it's a huge problem space. In general, I think we need to basically build from a zero-trust architecture. The thing is... the more functionality you want to give llm apps, the less secure they become. The is shown quite well by the fact that you can't render markdown images AND fetch content from anywhere within the same chat or there is a way to leak the chat history.