esperando o reset do Codex e farei um bench com o @OpenCodeReview_ comparando os reviews do Sol 6.2 x Sol 6.1 x Opus 5.5, vou ver se da pra comparar ambos no Medium e High também
já descobriram o que o sol 6.2 e o astra 6.2 são bons?
vou aguardar a galera descobrir para testar, minha bolha ta muito quieta sobre isso
1
1
115
fazer um navegador do zero é caro pra caralho, pqp.
só pra gerar as releases já preciso de runners parrudos, e a conta não perdoa
isso usando o @useblacksmith, que é mais barato imagina se estivesse rodando tudo nos runners do GitHub Actions 💸
22
"Linux is insecure shit."
That's not what the GrapheneOS Team was saying.
GrapheneOS was talking about THEIR ASOP/Linux problems they inherit, not arguing that Linux is shit, yeach, it is flawed, but that doesn't mean its useless. Quite the opposite, its an extremely capable os kernel, but that creates the problem.
Capabilities ≠ security.
Linux is a massive monolithic kernel, it contains a huge amount of privileged code, and kernel vulnerabilities are inevitable.
GrapheneOS inherits those problems from Linux thats why they patch the shit out of AOSP's Linux to maintain their security standards.
But the GOS team doesn't mean, that Linux developers are stupid, its a systematic problem that GOS is pointing out.
GOS team has discovered and reported vulnerabilities, contributed kernel hardening upstream, and helped patch many bugs and exploits over the years.
SELinux blocks cant help with EVERY kernel exploit,
GraphenOS assumes the kernel has vulnerabilities.
Thats why its designed for those vulnerabilities to be reached, and even harder to exploit.
>MTE,
>Kernel Hardnening
>Reduced Attack-surface
Linux is freedom that sets a foundation.
GrapheneOS improves on it for security.
Stay safe.
Get GrapheneOS.
Caio retweeted
só avisei o Chopin que tinha uma versão nova no @OpenCodeReview_
ele identificou nas releases, gerou uma imagem e passou pro Mozart fazer a postagem
além de melhorias vindas de contribuições, temos agora também integração com o Cursor
o Compose 2.5 fast foi incrivelmente rápido
OpenCodeReview 0.4.0 🚀
Review with Cursor
- Sign in to your Cursor plan in the browser, or add a user API key.
- The model picker offers the models your Cursor plan lists.
- Cursor accounts join priority and failover like the other providers.
1
3
184
finally
Replying to @asklumo
3/5
🔗 Connectors let Lumo work with your apps and files.
📅 Schedules let you automate recurring tasks.
✨ Skills let you personalize Lumo with reusable instructions.
This means you're spending less time on repetitive work, and more time on what matters.
16
aqui estou seguro bloqueando TLDs como esses com NextDNS
🚨 Yesterday, a major supply-chain attack used Optimizely to deliver malware through websites belonging to 178 organizations — including the BBC, CNN, Zoom and GitLab.
We conducted a forensic analysis of historical DNS telemetry and detection records to compare NextDNS with other security providers. The timeline below shows the earliest recorded detections, domain by domain.
NextDNS was far ahead of most providers on both coverage and early detection, by a margin that surprised even us. Our security heuristics and other advanced security features blocked domains used for malware delivery, payment data theft and iOS exploits.
Other providers were often hours or even days behind. Many still missed domains when we checked.
And now, back to our regularly scheduled Security Month programming.
57
blog.report-uri.com/worlds-l… is why diagrams.net doesn't inject third-party JS that isn't core to the app's operation.
Caio retweeted
It seems the clean up is done and the cat's out of the bag with who's responsible, so we've published:
blog.report-uri.com/worlds-l…
eu sei que @ProtonPrivacy não apoia OpenAI, Anthropic, Grok, etc, mas apesar disso tudo, deveriam oferecer integrações como essa do @1Password em todos os providers
em todo lugar só vejo 1Password..
We’re starting to test 1password in dots!
Our goal is to make it easy for your dot to do tasks that require a login, while keeping your credentials secure. Excited to roll this out more broadly soon.
Let us know what you think.
1
1
78
In January, almost every PR at Linear had a second human reviewer. By October, 53% were reviewed by an agent and the author alone.
We’ve become more deliberate about where human attention goes.
Here’s what we’ve learned:
linear.app/data/code-review
Caio retweeted
We're releasing a free new tool to vetted organizations to help detect child sexual abuse material, including AI-generated images. We're also adding ChatGPT warnings, publishing guidance for developers, and supporting child safety investigators.
We developed the classifier with assistance from the National Center for Missing & Exploited Children. We're working with the Tech Coalition and ROOST to help make it available to more organizations.
openai.com/index/strengtheni…
Caio retweeted
pessoar, to me inscrevendo em outras vaguinhas de sec (sou blue team) , vou deixar meu linkedin abaixo, caso tenham uma oportunidade e meu perfil bater, da um up aqui!! linkedin.com/in/cyb3rjulia
Caio retweeted
eu plantei bugs propositais em uma PR com 30 arquivos e coloquei coderabbit e @OpenCodeReview_ lado a lado para batalhar
github.com/bpinheiroms/refin…
o resultado foi:
- ambos não tiveram falso positivo
- opencodereview encontrou 2 bugs a mais que o coderabbit
- opencodereview terminou em 4 minutos ao invés de 18 minutos do coderabbit
o melhor de tudo: opencodereview é self hosted e tu pode usar teu próprio provider que já paga :)
Caio retweeted
finalmente Mozilla e Google aprovaram minha extensão que melhora a UI do github, quem quiser baixar, pode baixar direto da loja oficial:
> Mozila: addons.mozilla.org/en-US/fir…
> Chrome: chromewebstore.google.com/de…
Quem quiser contribuir:
github.com/bpinheiroms/refin…
a mente do pessoal ta corrompida kkkkkkkkkkkkk
Replying to @cyberash699
You're misrepresenting what we said without bothering to read was was written and the context for it. It's exactly what we described in the thread you're responding to. This new thread was also posted on Reddit so it's unsurprising people are replying without reading the content.
12
Replying to @Guptha_Gudivada
AI automates Terraform, manifests, pipelines and basic troubleshooting, but ownership of production outcomes stays with humans. When an AI-generated change breaks things, the team that approved and deployed it is accountable.
DevOps is not becoming obsolete. The repetitive work is shrinking while the need for architecture, risk judgment, complex incident response, security and system design grows. Engineers who adapt to directing and validating AI output stay essential.
Caio retweeted
I deleted all my SSH keys
blog.infected.systems/posts/…
Caio retweeted
Replying to @bilskirnir3124 @thsottiaux
You can't blame LLMs for the incredibly insecure and buggy status quo of large mainstream software, firmware and hardware projects. Linux has hundreds of of preexisting severe vulnerabilities being found and fixed every month. Linux already churned out endless slop without LLMs.





























