
Co-founder & CTO @ @MilgramAI • Author of bettercap, pwnagotchi, opensnitch, bleah, legba and a few other things.
Joined January 2009
- Tweets20.9K
- Following2.1K
- Followers48K
- Likes27.8K
Simone Margaritelli retweeted
just 4 days later. MySQL pre-auth rce.
The best comment I have seen about this:
"So apparently, we got an RCE in every layer of the entire LAMP stack in like a month? Yes, M is not Postgres here, but I guess MySQL/MariaDB will follow soon." 😂
news.ycombinator.com/item?id…
Simone Margaritelli retweeted
Pre-auth RCE in nginx is the bug every hacker dreams of finding. It almost never happens.
Today we’re disclosing nginx2root: five nginx vulnerabilities, and an exploit that takes an unauthenticated attacker all the way to root.
One of themhits the default config.
If you run nginx, patch now. Then read: nginx2root.com
CVE-2026-123007, CVE-2026-123634, CVE-2026-123633, CVE-2026-121810, CVE-2026-113425
Simone Margaritelli retweeted
holy shit this seems gigantic.
now just imagine the deepseek team doing this with the deepseek harness. we might unironically get frontier-level performance just like that
Introducing Qwen3.8-27B-pi-adapted 🤗
We adapted Qwen's model specifically for the pi coding agent using reinforcement learning and a LoRA adapter.
The results, measured as pass@1:
🚀 SWE-bench Pro (Hard 51): 29.4% → 62.7%
💻 Terminal-Bench 2.1: 39.5% → 55.8%
⚡ 55–62% fewer tool calls per task
This is the power of open models: you can adapt them to YOUR tools and YOUR use cases. Let’s build! 💛
Open weights: huggingface.co/harness-hub/Q…
Methodology: huggingface.co/blog/harness-…
Now do ARTEX 😈
Announcing harness adapters for @Alibaba_Qwen Qwen3.8-27B on Pi harness!
We're releasing a LoRA adapter that makes Qwen3.8-27B a much better and efficient agent behind Pi harness.
SWE-Bench Pro (Hard 51 subset) pass@1: 29% -> 63%
Terminal-Bench 2.1 pass@1: 40% -> 56%
Much more efficient: 55-62% fewer tool calls per task 🤯
huggingface.co/harness-hub/Q…
The era of harness-specific adapters is now on! More harness adapters to come in the next days 🧵
2
4
1,154
Simone Margaritelli retweeted
Can a self-hosted open model stand in for a frontier model as a coding agent on real work?
Qwen3.8-27B + pi LoRA (4×A6000) vs Opus 5.5, same bug fix in our inference proxy (lost request logs after logrotate).
Both green, ~90s each.
Local AI will win.
Simone Margaritelli retweeted
Security teams were told to make lemonade from whatever the year handed them. Nobody mentioned the chute.
Simone Margaritelli retweeted
Yo no sé qué pasa, que #Nintendo eShop está regalando #Skyrim si accedes desde tu cuenta Nintendo. Por si os queréis hacer con él antes de que parcheen.
Y si no es un bug, dadle las gracias a Bethesda!
offframe.com/en/the-nintendo…
#Java4Shell #CVE-2026-114938
Achieving a full RCE exploit chain is quite challenging. Heap memory disclosure, on the other hand, is relatively easy to achieve. But the real danger comes when sensitive data, like AWS access keys or something, happens to reside in nearby memory regions. This is reminiscent of #Heartbleed back in the day.
Simone Margaritelli retweeted
‼️ ALERT - Microsoft's AI Agent Framework package has been hijacked on PyPI.
Attackers pushed two malicious versions (1.26.1 and 1.26.2) that steal cloud credentials and tokens—silently, on EVERY Python startup.
Check NOW ↓ thehackernews.com/2026/11/at…
Simone Margaritelli retweeted
JUST IN: 🇪🇺 Europe's largest natural gas supplier 'Equinor' says it is under cyberattack.
Simone Margaritelli retweeted
BREAKING 🇳🇴: Norway energy giant is detecting and currently fighting a cyberattack attack to keep the power on
Simone Margaritelli retweeted
🚨We urge orgs to immediately patch the recently disclosed Oracle Java vuln CVE-2026-114938. This critical memory-corruption vulnerability, known as “Java4Shell,” affects the Base64 decoder of the x86-64 HotSpot JVM, part of the Java Development Kit. 🔗go.dhs.gov/5CC





























