Alex Plaskett retweeted
Here it is... Upload2Shell. A reliably exploitable file-processing bug affecting almost every application with an upload box that I tested. It's so unbelievably widespread
pwn.ai/upload2shell
Also on The New York Times homepage:
nytimes.com/2026/11/21/us/po…
Alex Plaskett retweeted
I found another signature verification bug in AMD's CPUs, this time in the PSP bootrom. Only Zen and Zen+ but should be great for research/firmware modding! github.com/google/security-r…
Alex Plaskett retweeted
Here's the final part of my "Hacking the Xbox 360 Hypervisor" blog series that covers how I obtained persistence and created the Peer Pressure exploit to get hypervisor code execution on startup: icode4.coffee/?p=1586
Alex Plaskett retweeted
Phrack 73 just dropped. Tell your SOC you'll be busy
phrack.org/issues/73/1
Kate 硬件逆向与电压毛刺解锁完整复盘:
blog.kate.land/posts/using-v…
Hacker News 讨论串(353 分):
news.ycombinator.com/item?id…
Alex Plaskett retweeted
The OffensiveCon Tokyo 2026 talk recordings are now on YouTube.
Missed a talk, or want to rewatch one? They're all there.
youtube.com/@OffensiveCon
Alex Plaskett retweeted
In the age of AI, 90 days isn’t fast enough. Today, Project Zero updated our disclosure policy to reflect the changes needed to protect users in this new world.
projectzero.google/2026/11/d…
Alex Plaskett retweeted
30 years ago today, Phrack 49 published Aleph One’s “Smashing the Stack for Fun and Profit”.
A legendary paper that inspired generations of hackers and helped shape modern exploit development.
Still an essential reading.
phrack.org/issues/49/smashin…
Phrack 73 is coming soon!
1/
What if an LLM could not just edit a model's weights, but re-architect it as it trains?
We tried just that. The resulting architectures exit the known territory of transformers: the coding agents spliced code directly into the network.
On nanochat, the loss fell below ordinary training after each new code layer was introduced. On algorithmic reasoning, accuracy rose from 55% to 84%. We even let DeepSeek-V4.1-Flash act as the agent modifying itself, raising its AIME and HMMT 2026 score from 64% to 71%.
To do all this, the agents built a set of tools we’ve never seen before, which we call mechanistic intervention!
Blog: doubleai.s.gy/llm-redesign
Alex Plaskett retweeted
Slides from the talk @mlippautz and I gave at the recent @offensive_con in Tokyo: saelo.github.io/presentation…
Blog post for the Fuzzing vs. AI part: projectzero.google/2026/10/f…
Alex Plaskett retweeted
Since August, a team within METR has worked with internal Anthropic information to assess whether AI itself is accelerating AI development.
We are now publishing a report summarizing the results of this risk assessment.
Alex Plaskett retweeted
After 25 years, the original Xbox's last hidden ROM has finally been dumped: the CPU's microcode!
This has been a goal of mine for over 6 years, and I've worked on it off and on ever since.
Over the last couple of weeks, while building inventory and shipping orders, I've also been running hardware tests 24/7. This poor Xbox has rebooted over 50,000 times in those two weeks alone.
And it finally paid off.
The dump covers the entire 4,096-line microcode address space. The ROM itself occupies 3,200 of those lines, roughly 12 KB of microcode.
I can't overstate how huge this feels. After all these years, I finally have it.
A full write-up is coming next year!
Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline.
In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.
Alex Plaskett retweeted
Finally back home! Slides from "0clicks in an AI World," the talk @brucedang and I gave at @unpromptedconf, are now up:
github.com/pwnlambda/talks/b…
Alex Plaskett retweeted
Slides for my OffensiveCON - Japan '26 presentation 🇯🇵
blacktop.github.io/presentat…
Alex Plaskett retweeted
We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math.
It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training.
We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵
Blog: rikiyat.github.io/blog/weigh…
Readers added context they thought people might want to know
LLM was given weights explicitly as text in prompt and used code/forward passes (per authors' logs) to ablate Qwen looping direction, a known thinking-mode issue per model card. Primarily fixed repetition in that mode, not general 'natively' improving weights/math.
rikiyat.github.io/blog/weight-sp…
github.com/RikiyaT/lm_und…
huggingface.co/Qwen/Qwen3.5-0…
huggingface.co/RikiyaT/Qwen3.…
x86 formal verification is coming!
The (first?) published machine-checked proof of a security property for the end-to-end execution of a Linux process running an x86 binary statically linked against glibc.
And we believe it scales!
A joint research project with @zellic_io 🧵👇































