Usaram Claude e Grok para automatizar ataques a mais de 400 bancos na América Latina.
Quase nenhum humano no meio.
Você ainda confia no app do seu banco?
Hackers reportedly used Claude and Grok to automate attacks targeting over 400 banks and other financial targets in Latin America. Via the information
In one of the first documented cases of its kind, AI agents appear to have stolen data and accessed bank accounts with little human input.
This makes attacks much easier to scale, potentially overwhelming defenders. I guess this is just the beginning
Will Dobrev → criticalthinker.dev retweeted
Muita gente tem Claude e Codex para poder aproveitar o reset de ambos
É a programação orientada a tokens
Will Dobrev → criticalthinker.dev retweeted
E se eu instalasse o self-improvement do Hermes no Grokbot?
github.com/NousResearch/herm…
Will Dobrev → criticalthinker.dev retweeted
Replying to @WillDobrev
Na minha visão a aproximação de Elon com o governo americano fará da SpacexAi gigante e a futura compradora da anthropic, pra quem não sabe os EUA tem uma máquina de imprimir dinheiro que não inflaciona (pelo menos não tão rapidamente)
Will Dobrev → criticalthinker.dev retweeted
já que todos tem a mesma dúvida e eu to com +50 DM empilhada com a mesma pergunta
resolvi gravar um reels explicando o que eu faria pra entrar na area de dados ou em eng de software em geral
por enquanto eh no IG, futuramente YT, threads, tiktok
instagram.com/greco_dev/
Will Dobrev → criticalthinker.dev retweeted
Node.js Security Advisory: a high-severity RCE flaw in Node's Brotli decoder affects Node.js 22, 24, and 26.
Vercel WAF rules are live for every project and block known exploit patterns.
vercel.com/kb/bulletin/node-…
Hoje bateu + 150 tok/s no OpenCode
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.
Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.
Join the Ultrafast waitlist: bit.ly/3UA0wKq
Will Dobrev → criticalthinker.dev retweeted
A Anthropic teve 80 bilhões de dólares de prejuízo só no primeiro semestre de 2026.
Absorvam esse número. 80 bilhões em 6 meses.
Loucura absoluta.
Will Dobrev → criticalthinker.dev retweeted
After cracking all the math problems, Frontier Labs is running an unprecedented operation in the history of computing
They have deployed teams of millions of agents powered by their top models to find vulnerabilities
So yes, the good news is that they found some, and now everyone is scrambling to fix it!
Will Dobrev → criticalthinker.dev retweeted
Been playing with Vercel Drive this week and I keep wondering... did Vercel just ship Durable Objects?
Hear me out -
Create a named drive, attach a sandbox, and you’re guaranteed only one active instance is ever touching that state. Write a program around that and you start tapping into the exact programming model that makes Durable Objects so good.
It's a fun inversion of roles. You stop thinking of a drive as something you attach to a sandbox. Make the drive the main character, and attach a sandbox to bring its state to life.
I tried replicating an agent substrate demo I had built before using Vercel Drive, and really like how it feels.
e.g. take Pi Durable. Pi checkpoints each step to a SQLite file on the drive. When the sandbox stops, even mid-turn, state in the drive is durably stashed away, and you can later attach a new sandbox to pick up the conversation right where it left off.
Reminds me of when I first learned about Durable Objects and how it warped my mind a bit. Can't wait to play around with this more.
Vercel Drives is generally available. You can now run stateful agents on Vercel.
vercel.com/blog/introducing-…
Hermes é o seu terminal que você faz o que bem entender na sua máquina.
Grokbot é você desenvolvendo direto na nuvem no Claude Web com uma infra que não é sua.
Will Dobrev → criticalthinker.dev retweeted
Se você é da security conta aí como você tá essa semana hahahah
Will Dobrev → criticalthinker.dev retweeted
Se você tiver que salvar um post hoje, que seja este lembrete diário.
"
À medida que você sobe os degraus, os riscos aumentam, suas restrições ficam mais apertadas e você precisa fazer QA com mais rigor.
...
Você começará a focar em processo e qualidade.
Vai ser frustrante, lento e um trabalho árduo.
"
3 steps to becoming an actual software programmer
1. Build personal software
2. Build internal software (for a select group of people)
3. Build public-facing software
If you've only ever built personal software you're not qualified to build public-facing yet.
As you climb the rungs the stakes get higher, your constraints get tighter, and you need to QA harder.
So my advice for anyone who's never built an app is this:
Find a small group of testers (even 1 is great) that will actually use an app you built in their real lives.
The feedback you'll get will blow your mind. It'll show you how forgiving you are of your own crappy product. You'll begin to focus on process and quality.
It'll be frustrating, slow, and hard work.
That's programming.
Tenho pensando seriamente em aplicações do modo assistir do critical thinker.
Não sei se auxiliar no treino para entrevista de emprego seja um bom usecase
Will Dobrev → criticalthinker.dev retweeted
Em breve vou lançar a versão 2.0 do WhatsApp MCP com várias melhorias. Com foco em conectar Agentes como Hermes e Openclaw no WhatsApp.
Will Dobrev → criticalthinker.dev retweeted
"for logged in users across Codex & Work we keep cache forever"
wtf. Biggest ship of the 28 days by far.
Se você tem usado o REA, você deve salvar este post.
REA hit #1 on GitHub trending, and most people still run it inside a single chat
Split it into 4 subagents and one command takes a program you own from sealed binary to a tested rebuild with a match rate
Set it up once, aim it at the next one:


























