threatactor.club

Teaser for my @bsidesfra presentation next week:

LLMs are kinda "cute" they create files like fix_final_perfect.go and then the code fails of course.

I don’t find the bugs - the bugs find me!

Let’s call it „beg4shell“. At least now everyone can be a vulnerability researcher.

Sooo Leute was sind denn eure Boomermeinungen zu https://www.langenscheidt.com/jugendwort-des-jahres bitte?

Forget wordle:

joernchen scored 0.06 which is amazing!
★★★★★
I've played 5 days in a row
https://beatmatch.pro/

Send memes my way please!

Niemand: ...

Deutsche JSON API:

{"status":"BAD_REQUEST","timestamp":"04-07-2026 05:38:22","message":"Missgebildete JSON-Anfrage".....

Re: https://infosec.exchange/@bsidesfra/116821688667636012

Soooo what would you like to hear from me in that talk?

Me: you Claude build me a fuzzing harness for this codebase

Claude: sure thing, here we go

Claude does build something with some back and forth with the compiler and linker and figures how to solve all the errors.

Claude: Now let me do a clean build….

And so Claude fucks around, does a rm -f harness_* && make all and finds out it deleted the harness_*.c files as well.

¯\_(ツ)_/¯

Putting on my tinfoil hat I would say that it’s in the best interest of Anthropic because well when it has to recreate the code again more tokens are being sold.

So many calcs so little time.

Special thanks to @buherator, best sticker I had in a while. 😍😍😍

https://badhost.org it's really bad. :P

I miss shitposting :(

Bling bling ;D

You use Claude Code to find vulnerabilities, I find vulnerabilities in Claude Code.

https://0day.click/recipe/2026-05-12-cc-rce/

https://greptalks.ai/ is fun, also https://greptalks.ai/rate-my-talk/ for letting it rate your own presentations.

Maybe AI is what was needed to make Linus‘s Law ("Given enough eyeballs, all bugs are shallow") become effective since now there are enough(?) AI assisted eyeballs.