data + ai + infosec engineering. @squirrelscan_ @opennem. at https://teapawt.net/t.co/yOGtW7uHIA and https://teapawt.net/t.co/kvkcg5TrG9
Australia
Joined March 2007
- Tweets1.7K
- Following762
- Followers458
- Likes4.4K
github.com/thecybergeek/php2…
php2shell OOM in mod_php (CVE-2026-109618): the fake-HashTable fire only runs ~14-byte commands, too short for a shell. Poison a PHP session via upload_progress with <?php system(base64_decode())?>, then exec in 11 bytes: php /tmp/s* www-data shell. Full PoCs in my github repo!
#php2shell #CVE #InfoSec #PHP #RCE
The Saudis?
FBI arrests Saudi OpenAI engineer over covert recording device, DOJ says reut.rs/4ztLULn reut.rs/4ztLULn
two claude sessions in separate accounts working on similar data and training tasks on my machine saw each other in process lists and one of them asked if its ok to create a shared directory so that they could send each other messages to coordinate what they have in common 🤷♀️
interesting insight into where RL effort is spent on these models - two-thirds is code
Nearly half a year of silence. We spent it studying one problem: how far RL can scale.
MiMo-V2.6 is in the middle of its RL run right now. Three things we scaled: compute (~2B tokens per step, 1568 prompts × 16 rollouts, fully async), environments and harnesses (multi-task agentic RL, mixed across multiple harnesses in one run), and grader compute (agentic in-group credit assignment, with test-case and rubric-based rewards). We'll open-source the details piece by piece over the coming weeks.
Streaming the run: mimo.xiaomi.com/rl/
