R&D lead for Xint Code (@xint_official). Flag capturer at @PlaidCTF. Cryptography enthusiast.
Seattle area
Joined September 2013
- Tweets693
- Following433
- Followers3K
- Likes2.1K
Pinned Tweet
Stealing OpenAI Reasoning Traces (October 2026 Edition)
Using a new technique called Compaction Smuggling, we could bypass all mitigations and steal reasoning traces from frontier OpenAI models.
🚨🚨 Xint researcher @tjbecker found a vulnerability that enabled large scale OpenAI model reasoning distillation that bypassed earlier mitigations set in place post-Stolen Thoughts. We disclosed responsibly to OpenAI and they completed patching last week. Full technical writeup here.
xint.io/blog/compaction-smug…
Tim Becker retweeted
From our work on Ghost Compact, we learned that compaction behavior is highly steerable. We wondered if we could convince a frontier model to transcribe its raw reasoning into the compaction, then use the compaction to smuggle it to the weaker model. The answer was yes!
Tim Becker retweeted
Need response concise explain table first diff6 second10 third7 sum23. verify align yes.
Tim Becker retweeted
An internal model at OpenAI has produced a claimed refutation of ETH! Russell Impagliazzo, Mohan Paturi, and I have been asked to help organize an open and independent examination of this claim. 1/4
From our work on Ghost Compact, we learned that compaction behavior is highly steerable. We wondered if we could convince a frontier model to transcribe its raw reasoning into the compaction, then use the compaction to smuggle it to the weaker model. The answer was yes!
For our research, OpenAI awarded @xint_official with a $6000 bounty. On October 23rd, OpenAI rolled out a new batch of mitigations intended to make reasoning traces harder to leak.
Tim Becker retweeted
I haven't formalized this anywhere, but I think the next 12-18 months will have an enormous amount of innovation in the cybersecurity industry/community.
I think there are some good reasons to think a number of those interventions will shift economics in favor of defenders, even against model capabilities in 12-18mo.
But there's a diffusion lag time and tons of frictions. Quite concerned about the "cyber poverty line" and the long tail of e.g. critical infrastructure.
Tim Becker retweeted
It’s going to be really funny if everyone pointing their LLMs at the PHP codebase tonight finds dozens of OTHER critical vulnerabilities aside from the ones getting fixed tomorrow
Tim Becker retweeted
CLAUDE FOUND THE SEEDS
anthropic.com/research/ecdsa…
Tim Becker retweeted
Hackers get $1,262,000 for 98 zero-days at Pwn2Own Ireland
bleepingcomputer.com/news/se…
bleepingcomputer.com/news/se…
Tim Becker retweeted
“I ran <best internal model> for N trillion tokens at <frontier lab> and it couldn’t even dent <fundamental assumption>” should automatically get Best Paper at CRYPTO/Eurocrypt.
Tim Becker retweeted
Had a blast competing as part of team @xint_official! Proud that all our exploits succeeded on the first try!
As always it was great meeting old and new friends! Until next time!
Congratulations to Team Xint (@yves_bieri @cdor1 @tjbecker @taisic_) for their 2nd place finish + $240k in winnings at #Pwn2Own Ireland:
🎯 5/5 on their exploit submissions, including the Home Assistant Green, Google Pixel 10, Oracle Autonomous AI Database, LiteLLM, and Philips Hue Bridge Pro
📲 First team to hack the Pixel 10 (and might have won us the whole thing if it didn't turn out to have been a known bug)
🤝 First time we competed as team Xint
And special thank you to all the organizers and judges at @thezdi for their work making such a high quality event happen and bringing together the best talent from around the world to compete
And of course congrats to @ikotas00 on their first place finish (first Japanese team to win Master of Pwn!)
