0K

@immunefi All Stars | A carefree cyber sailor. Solves security challenges. Secures protocols.

[0K @ZeroK_____]

One universal truth:

every codebase has Bugs waiting to be caught.

[0K @ZeroK_____]

@0x3b33 I got some of most interesting offer, i missed them due to X dm lol :)

[0K @ZeroK_____]

Educate yourself more and more. In the coming years, most people will ask AI to perform the most basic tasks for them instead of using their own minds to think. Have you ever heard the quote that says:

“Many People Would Rather DIE than Think
”?

This will be the case for 80-90% of people in this space in the upcoming months and years. So, educate yourself, sharpen your skills, and learn more about AI, programming, or anything related to your field so that you can be among the 10-20%.

AI is both a blessing and a curse at the same time.

[0K @ZeroK_____]

Well said

[pessimist @0xpessimist]

Listen white hat,

Hackers try to cheat systems through deep thinking, they find exploits, but they don’t try to cheat thinking itself. Many of you have had your brains hacked: your attention spans shortened, your thinking is outsourced. You’re left with nothing.

But you're building an AI harness, and you want it to find every bug in the world. Turns out, to build a solid AI harness, you need weeks and months of deep thinking before even taking the first step. But some of you have already given up that capability.

[0K @ZeroK_____]

@0xpessimist Yep :(

[0K @ZeroK_____]

Pro tip:

Never run your ai agent on a codebase you don’t understand!

[0K @ZeroK_____]

@AnticsDecoded based on real experience i had, this is a good way to use ai if you know what you're doing

[0K @ZeroK_____]

@victorokpukpan_ nothing, but you might miss valid bugs your ai says its invalid or vice versa.

[0K @ZeroK_____]

@lonelysloth_sec 🫡🫡🫡

[0K @ZeroK_____]

IMO, if you don't understand how a codebase works and why it works the way it does, then every finding can look either valid or invalid to you, because you have no real understanding of the system your AI is analyzing.

This becomes even more important when running AI on BBPs. For contests, you might get 5 valid findings out of dozens of invalid ones, and without sufficient domain knowledge, it's difficult to separate signal from noise.

Take Geth-like codebases as an example. If you already understand how Geth works, its architecture, invariants, and execution flow, then you're in a much better position to use AI effectively on codebases derived from or similar to Geth. The AI can help accelerate your research, but your understanding is what allows you to validate whether a finding actually matters or if even a one can exist or not.

[0K @ZeroK_____]

@0xTimofey Check dm brother

[0K @ZeroK_____]

@0xTimofey @immunefi @MitchellAmador 🙏🙏🙏

[0K @ZeroK_____]

@cantinasecurity @injective Launching a bounty for a group of scammers is crazy!!

[0K @ZeroK_____]

@adrianhetman Me

[0K @ZeroK_____]

Back to beginning again, Every time I think the AI is ready, I realize it’s not quite there yet, and it definitely needs some more work!

[0K @ZeroK_____]

@m4rio_eth What you mean by saying threat modeling engagement please?

[LoopGhost @LoopGhost007]

🧵[1/6]

Last week I privately disclosed a valid vulnerability affecting Alephium’s Wormhole bridge integration.

The issue was confirmed by the team. It allowed a permissionless attacker to drain the entire balance of a live mainnet contract.

[0K @ZeroK_____]

@lonelysloth_sec which model is best then ? gpt 5.5 ?

[0K @ZeroK_____]

And this AI is not meant to spam reports. My goal is not to make projects waste time and effort reviewing non-worthy findings.

The goal is simply to find bugs that might otherwise go unnoticed. Every finding should go through at least 10 verification stages, followed by another review strong step before anything gets submitted.

I’ll be incredibly happy if it help prevent even 1 out of every 10 hacks we see nowadays.

Also, this is not marketing. I have zero plans to sell AI products, or start a business around them, or make noise about what I’m building. I’m still a complete beginner in the AI space, and I’m just doing my best to build something that works well for me.

So no marketing BS here and in future posts either 😂

[0K @ZeroK_____]

The AI agent I’m currently creating is not meant to replace me or any other security researcher.

The goal is to build something closer to a senior SR, one that can work faster than me and surface exploit paths that might take a human more time to find.

But I don’t believe it will ever fully replace the intuition of top tier hunters, at least not for now and upcoming year!. Maybe in the coming weeks or months it might catch some strong critical's with little to no human involvement lol, but for now, I see it as a powerful agent that can work on many bbp in same time, not a replacement.