Clojurians Log v2

can we SEE the model? > no

I guess, if you don't like AI code, and an LLM sends you a patch for a zero day, what do you do? Gotta do something

Obviously. I just don't buy the recent anthropic FUD who are just trying to grift more govt contract money

Even ffmpeg is taking patches from Anthropic.

The marketing antics are wild lol. It reminds me of the OpenAI Q* debacle

I'm mostly looking forward to the hardening that will take place. People talk like it's the end of the world... Once things are finally secure, they'll be secure. Then we won't have to worry about models hacking everything.

Things won't be secure

You assume code and infra environment is static, but it's constantly in flux and reintroducing vulnerabilities, especially as more and more of it is gen coded 🤣

SQL Injection is a very simple issue to detect, even more to fix. Essentially, a solved problem. My wife's middle name contains an '. An insanely huge number of sites don't accept it... So no, I don't believe we'll ever be "secure".

Bobby tables

They’re gearing up for an IPO I believe?

This is just about government contract imo

Reminder

all software is insecure until proven otherwise i.e. with https://lean-lang.org