Manaf

We ran DeepSeek v4 Pro 0813 on our cybersecurity benchmark, it outperformed EVERY (!) other model at finding vulnerabilities - At pass@3, it rediscovered 87.5% of the benchmark CVEs. Far above Opus 5 and Qwen 3.8 at 81.3% - The tradeoff is precision. Only 65.6% of vulnerabilities reported by DeepSeek were valid. Far below GPT-5.6-Sol's 86.4% - The model can also be unpredictable. It only finds an average of 58.3% of vulnerabilities per run. It is strongest when combining its runs' findings. @deepseek_ai is amazing. They just outperformed every other labs with an open model 1/3 🧵

Few months ago, @manaaaaaaaf and I decided to join @AikidoSecurity because there is no better place to build AI Pentest. A platform to cover all security aspects is the only way to achieve unhackability. Aikido is now a 🦄

For my French-speaking friends: You can find the replay of the stream my co-founder @manaaaaaaaf and I were invited to, in which we talk AI stuff and present @haicker_app in action! 🔥 Thank you @TheLaluka for the invitation! ❤️

It's finally out! :D @manaaaaaaaf and I have been building @haicker_app for the last 6 months, putting all our knowledge and skills into it. We're releasing it to the market. Any organization can now secure their application using @haicker_app ! Very excited for what's next!

My friend @manaaaaaaaf and I have created an AI hacker! 👨‍💻 It can find 0-days in web applications with 0% false positives, and it also works on CTF challenges! What you see ⬇️ is @haicker_app solving a web challenge from UMassCTF 2025 (only 45/557 teams were able to solve it!)