RSSAmplifier

Blog

dolftax

Recent content on dolftax

dolftax.comRSS feed ↗2 posts

Latest posts

LLMs are great at finding bugs, terrible at finding them twice

Run an LLM-based code reviewer on a code change 10 times and it flags a SQL injection vulnerability 7 times. The other 3 runs come back clean. Same code, same vulnerability, different result. 
 LLMs are probabilistic, but security requirements are binary. “Usually catches security issues” is a bug, not a feature. 
 When LLMs do catch something, the finding is often good because…

About

Hi, I’m Jai [/dʒeɪ/]. 
 I write about software engineering, security, systems design, longevity, and the intersection of tools and thinking. 
 Blog - longer essays and explorations. 
 Notes - quick thoughts and observations. 
 Currently building DeepSource . 
 You can find me on Twitter or reach out via email .