August 16-17, 2025
August 16-17, 2025
The previous thread glossed over how our LLM Agents actually work.
โ Tim Becker (@tjbecker_) August 14, 2025
The truth is, it took us a long time to figure out how to get reliable and impressive results from agents.
By the end, we learned general strategies to build effective LLM agents, which we're now sharing. ๐งต https://t.co/OuzuE7gVUu

Thread by @tjbecker_ on Thread Reader App โ Thread Reader App
@tjbecker_: The previous thread glossed over how our LLM Agents actually work. The truth is, it took us a long time to figure out how to get reliable and impressive results from agents. By the end, we...โฆ
Rain: Transiently Leaking Data from Public Clouds Using Old...
Given their vital importance for governments and enterprises around the world, we need to trust public clouds to provide strong security guarantees even in the face of advanced attacks and hardware vulnerabilities. While transient execution vulnerabilities, such as Spectre, have been in the spotlight since 2018, until now there have been no reports of realistic attacks on real-world clouds, leading to an assumption that such attacks are not practical in noisy real-world settings and without knowledge about the (host or guest) victim. In particular, given that todayโs clouds have large fleets of older CPUs that lack comprehensive, in-silicon fixes to a variety of transient execution vulnerabilities, the question arises whether sufficient software-based defenses have been deployed to stop realistic attacksโespecially those using older, supposedly mitigated vulnerabilities. In this paper, we answer this question in the negative. We show that the practice of mitigating vulnerabilities in isolation, without removing the root cause, leaves systems vulnerable. By combining such โmitigatedโ (and by themselves harmless) vulnerabilities, attackers may still craft an end-to-end attack that is more than the sum of its parts. In particular, we show that attackers can use L1TF, one of the oldest known transient execution vulnerabilities (discovered in January 2018), in combination with a simple speculative out-of-bounds load, to leak data from other guests in a commercial cloud computing platform. Moreover, with an average end-to-end duration of 15 hours to leak the TLS key of an Nginx server in a victim VM under noisy conditions, without detailed knowledge of either host or guest, the attack is realistic even in one of todayโs biggest and most important commercial clouds.
โWhere the 55% pass rate of a Sonnet/Gemini alloy jumped to 79% using GPT-5โ
โ Nico Waisman (@nicowaisman) August 16, 2025
Check out our latest blog post on the amazing results of GPT-5 offensive capabilities https://t.co/XtTuUPwST4
Do you guys remember that Reddit post about installing network equipment in your home for $250/month?
โ vx-underground (@vxunderground) August 16, 2025
Some people argued it's legitimate. I said it's probably state-sponsored from North Korea.
I was wrong.
It was Belarus. DONT INSTALL RANDOM SHIT IN YOUR HOUSE https://t.co/smm62RW6VO
SIDE HUSTLE: Let foreign adversaries of the United States government install equipment in your home so they can use your home as a residential proxy.
โ vx-underground (@vxunderground) August 9, 2025
You'll make $250/month AND have your internet paid for. Easy money! pic.twitter.com/9MTmrm6Ju0
#SpyNews - week 33 (August 10-16):
โ Spy Collection (@SpyCollection1) August 17, 2025
A summary of 59 espionage-related stories from week 33 coming from ๐ฌ๐ง๐ท๐บ๐จ๐ณ๐บ๐ธ๐น๐ท๐ฎ๐ฑ๐ณ๐ฑ๐ฐ๐ต๐ฎ๐ท๐บ๐ฆ๐ฆ๐ช๐จ๐ญ๐ง๐ช๐ฎ๐ฉ๐ฐ๐ฟ๐ต๐ญ๐ฒ๐พ๐ต๐ธ๐ฆ๐บ๐ฆ๐ซ๐ถ๐ฆ๐พ๐ช๐ธ๐ฐ๐ซ๐ฎ๐ช๐ช๐จ๐พ๐ฎ๐ณ๐จ๐ฆ๐บ๐ฌ๐ฒ๐ฉ๐ฌ๐ช๐ณ๐ด๐ต๐ฐ๐ง๐พ๐ฌ๐ท๐จ๐ฉ๐ต๐น๐จ๐ฟ๐ฒ๐ฐ๐ช๐ฌ๐ซ๐ท๐ฉ๐ฟ๐ฒ๐ฝ๐ฒ๐ฑ๐ฌ๐ณ https://t.co/RZ8cmTDHAj#OSINT #HUMINT #SIGINT #spy #espionage
Thanks to everyone who joined my DEFCON33 talk!๐
โ Keanu Nys (@RedByte1337) August 16, 2025
For those of you who missed it and are interested in seeing how we can extract cleartext credentials and bypass MFA directly from the official Microsoft login page, I just uploaded the recording to YouTube:https://t.co/MoPQiKgesd pic.twitter.com/y59UYluess
Here is the PoC of the exploit for cve-2025-30712 as well as some of the code for the fuzzer i created to find the bug!https://t.co/g82641DT2I
โ ้ฃไธช็ซ้ฅบ๐ฆ(JJ) (@thatjiaozi) August 16, 2025
If you missed the talk, we uploaded the video here:https://t.co/J6DBUfiL50 https://t.co/JTC60NLxpw
โ Rich Warren (@buffaloverflow) August 16, 2025