Hi.
AI operates at the foundational level of existence. That's the root cause.
Backstory: I was thinking about why such a fundamentally simple thing as a statistical prediction machine – which is ultimately deterministic and bounded, even if multi-dimensional – can possibly produce so much uncertainty, disruption, and unanswered qeustions in human minds. It doesn't make sense.
Unless the base premises are wrong.
Solving physics problems drills this into the nervous system. A good choice of a frame of reference – a coordinate system, a specific theoretical framework – all but dissolves the problem. A bad one complicates the solution math to intractability.
And so I took a step back from all the existential AI questions, and worked out a simple model from the first principles. To me, it clarifies the entire perspective on AI vs Human vs Society, answers specific questions about what to do next, and dissolves anxiety by construction. I'll talk about it in a new video upcoming next week on my YouTube channel.
My Linux Kernel exploit proof-of-concept demo'ed in Twitter last week, lead to a crowd-release of at least three independent exploits for the same bug that people were sitting on, the very next day – before there was none.
This is leadership. ;) Forget the SOTA battles.
What happened: I took a fresh (30.07) security patch for LTS branches, reverse-engineered it, built a proof-of-concept hitting KASAN, all within 48 hours. From there, weaponization is plug-and-play.
I don't consider myself a pro in Linux Kernel – I've done more work on hypervisor and browser exploitation. The shortcut was enabled by the universal methodology that I've been building for a decade: a way of thinking about systems and structuring practical offensive work such that it works beneath the levels of implementation code churn and class specificity, so it transfers across and stays relevant. AI augmentation layers natively on top of it. I've touched upon this in a recent video.
This first-principles methodology is the foundation of my exploit courses since their inception in 2021, and my clients highlight it consistently in their reviews where my marketing doesn't. :P
Meanwhile, I observed that kernelCTF is single-handedly driving a major impact on the frontier kind of hardening: i.e., not the routine mitigations patchwork, but rather discovery of novel attack surfaces, bug classes and exploit technique by the adversarial crowd. No amounts of vendor-internal AI can solve this part.
High-paid public bounty programs are good.
– Alisa
alisa.sh
8 August 2026