# Machine Learning and Blockchain Security

Source: <https://schnack.dev/>

Author: Lucas Schnack

research. writeups. stray thoughts.

### European AI [europe2031.ai](https://europe2031.ai/)

Paralysis.

### Fat Tails [fooledbyrandomness.com](https://www.fooledbyrandomness.com/) · [en.wikipedia.org](https://en.wikipedia.org/wiki/Fat-tailed_distribution)

distributions with large skewness or kurtosis compared to the normal or exponential distribution

The world is normal until it isn't.

### Red or blue button? [x.com](https://x.com/waitbutwhy/status/2047710215265730755)

blue for sure

### Information Geometry [en.wikipedia.org](https://en.wikipedia.org/wiki/Information_geometry) · [franknielsen.github.io](https://franknielsen.github.io/IG/index.html)

A beautiful framework connecting probability theory and differential geometry.

It allows for intuitive geometric interpretations of relationships between probability distributions.

### WebAssembly [webassembly.github.io](https://webassembly.github.io/spec/core/)

Bytecode for a stack machine not only used in browsers but also in blockchains.

Countless ways to break things.

### Optimal Transport [en.wikipedia.org](https://en.wikipedia.org/wiki/Transportation_theory_(mathematics))

How can we morph probability distributions into others?

Everything is a flow.

### AI Agent Security [github.com](https://github.com/anthropics/claude-code/security/advisories/GHSA-5hhx-v7f6-x7gv)

LLM CLI tools can also be exploited — by AI agents?

hack claude make no mistakes

### AI Alignment [en.wikipedia.org](https://en.wikipedia.org/wiki/AI_alignment) · [ifanyonebuildsit.com](https://ifanyonebuildsit.com/)

How can we steer AI according to our values and goals?

Alignment is hard.

## ABOUT

Computer Science M.Sc. student at TUM.

Machine Learning with a focus on privacy-preserving methods.

Breaking blockchains (and other stuff).

## LATEST

2026-09-17 [AI Security](https://schnack.dev/blog/ai-security/), [Security](https://schnack.dev/blog/security/), [AI](https://schnack.dev/blog/ai/)

## [Containment Failures in AI Coding Agents](https://schnack.dev/blog/ai-security/ai-coding-agent-containment/)

Six vulnerabilities in Codex and Gemini CLI reveal how startup, sandboxing, configuration, and approval checks in coding agents can fail.

2021-01-23 [Kaggle](https://schnack.dev/blog/kaggle/), [AI](https://schnack.dev/blog/ai/)

## [Layers of Rock Paper Scissors](https://schnack.dev/blog/kaggle/rock-paper-scissors-ensemble-meta-strategy/)

Post-mortem analysis and strategy writeup for the Rock Paper Scissors competition on Kaggle.

[Read the blog →](https://schnack.dev/blog/)

## CONTACT

### \> SOCIAL

- [MAIL::contact@schnack.dev](mailto:contact@schnack.dev)
- [X::0zSchnack](https://twitter.com/0zSchnack)
- [LINKEDIN::lucas-schnack](https://linkedin.com/in/lucas-schnack)

### \> PROFILES

- [GITHUB::0Zeta](https://github.com/0Zeta)
- [HACKENPROOF::0Zeta](https://hackenproof.com/hackers/0Zeta)
- [IMMUNEFI::Zeta](https://immunefi.com/profile/Zeta/)
- [BUGCROWD::0Zeta](https://bugcrowd.com/h/0Zeta)
- [KAGGLE::luji01](https://www.kaggle.com/luji01)