Pavursec
I hosted an AI debate tournament! GPT 5.2 dominated with an impressive 85% win rate and the tournament taught me a lot about how LLMs think and interact. Some of the most interesting insights: There are meaningful and measurable differences between LLMs in terms of their debate skill. Debate skill correlates with other LLM benchmarks - generally “better” models are better debaters. LLM…
Static == Secure Right? If you’ve read my recent post extolling the virtues of static site generators for secure web development you might think that deploying a static site makes you more or less invulnerable to cyber adversaries. Although it’s true that going static dramatically reduces your vulnerability profile, it doesn’t quite eliminate it. This post considers five possible…
Juggling Priorities It should come as no surprise that, when building this site, my first concern was security. This arose partially from the fact that I live and breathe cybersecurity and believe (not without cause) that the whole internet is out to get me. However, it’s also true that sites focused on cyber issues tend to present attractive targets to hackers attempting to prove their…