Xinyang Ge (@aegiryy) profile picture
Joined October 2009
Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)Photo by Xinyang Ge (@aegiryy)
JUST IN: Riemann Hypothesis has been solved. alpo.ge/remark.pdf
1,820
8,108
5,369
61,192
15,287,576
Xinyang Ge retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
61,457
121,815
24,530
1,175,716
59,698,604
we should do both: keeping the current software patched while rewriting the whole world via principled designs. the fundamental paradigm shift is that code gen is getting cheaper and cheaper, so we can afford doing things in ways that aren’t imaginable before
So much software out there is just plain unpatch-able. It's going to be better for everyone to start new, even if it means a migration, than to keep trying to throw patches onto unsafe or insecure or outdated designs.
1
7
1,237
time is being compressed in the ai era
Anthropic is making some important changes to our Coordinated Vulnerability Disclosure policy in order to keep pace with the rapid improvement of AI driven vulnerability discovery capabilities. Starting in January we are moving to a 30-day timeline for Critical and High vulnerabilities, and 45 days for everything else. See the post linked below for more details. We know that this change may put more pressure on OSS maintainers and vendors. Given this change some important aspects of our policy are worth calling out here: * We will continue to agree to a sustainable pace with maintainers and vendors before sending a large volume of findings. * We will always include a patch or mitigation with each finding where possible. * We will work with vendors and OSS maintainers on a timeline for issues that can’t be fixed quickly. * A reminder that issues reported with no human triage, such as those from OSS-Scanner are not subject to this policy. See our FAQ for more details. We know that we cannot just patch our way to better security. Securing all of the code in the world requires a combination of language modernization, redesign, rewriting, hardening, reducing attack surface, and yes, finding and fixing vulnerabilities. But we must move at a faster pace together in order to make progress. anthropic.com/research/coord…
7
929
it matches my experience that the chance for another llm to find the same vulnerability within a short period of time is high
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…
1
1
1
24
3,620
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…
30
140
51
559
150,843
Xinyang Ge retweeted
Replying to @knwldgdigger
most are critical/high (should’ve posted the severity breakdown) i’ve intentionally held non-crit/high reports to help google prioritize
3
1
14
861
100+ chrome cves by a single researcher would've been unimaginable in the pre-llm era. what a year for ai-driven vulnerability research (and we still have two months left in 2026)!
8
10
2
139
11,118
if you're interested in building next-gen sandboxing for agents, have a ton of passion/ambition for low-level systems software (OS/hypervisor), and are itching to rewrite their internals for security/performance/whatever, don't miss this: job-boards.greenhouse.io/ant…
22
35
4
589
55,593
yes, i can't think of a more severe bug than being able to root a remote server w/o the victim even having a chance to respond. patch CVE-2026-98432 (or disable mptcp in the meanwhile)
I think this is the worst bug in the history of the internet Please patch ASAP everyone
4
17
5,108
claude found this one and built a 0-click rce chain: it takes 48min to leak a kernel pointer and defeat kaslr remotely, and another 5min to hijack the control flow and root the server over the network. patch it now (or disable mptcp).
AWWWWOOOOOOOGAAAAAAA 9.8 severity use-after-free vulnerability in the kernel’s IPv4 TCP stack just landed Time to patch! ubuntu.com/security/CVE-2026…
9
14
3
172
26,000
hit #1 on the chrome vrp leaderboard (with zero bounty) 😂
👀👀 we have a new 2026 Chrome CVE leader! chrome.def1ant.xyz/
4
2
83
5,544
Xinyang Ge retweeted
Claude found with ~zero prompting: - Linux kernel 0click "packet to root" RCE (!!!!!) - Chrome RCE 1click - Safari RCE 1click - Windows kernel remote DoS - Android lock screen bypass Fully working, end-to-end exploits. Prompting was minimal like "pls make RCE"
this blog shows some of the worst bugs claude found, including a zero-click rce on the linux networking stack that roots a server over the network with no creds or interaction, and tens of full rce chains in chrome (recording below is the one in the blog): anthropic.com/research/disco…
34
102
17
1,405
95,768
this blog shows some of the worst bugs claude found, including a zero-click rce on the linux networking stack that roots a server over the network with no creds or interaction, and tens of full rce chains in chrome (recording below is the one in the blog): anthropic.com/research/disco…
9
52
10
504
104,413
claude is able to find critical vulnerabilities in closed-source binaries as if they were source code (more on the details later). the gap between registers/instructions/memory bytes and its source code representation isn’t as large to llms as it is to humans.
9
2
81
8,303
my intuition is that it’s all control flows and data flows whether it’s source code or binary, and llms can just reason about the graph wherever it is from. plus llms don’t accumulate fatigues like humans do. i’ll share vulnerability details after they are patched.
1
12
1,667
Xinyang Ge retweeted
636606729769440499166579950236036751749912014371509557713570027508971809534551913252252094954941974952859310861988904737359709200557919 is a factor of RSA-896 saweis.net/posts/rsa-896.htm…
220
984
323
9,450
3,921,691
Unikernel is still pre-codified but just linked against the application into a single binary. We need JIT-kernel, in which LLM just-in-time generates an OS kernel tailored for the application it builds.
Fully agree with Zhihao. As the cost of implementation and verification asymptotically approaches 0, there is little argument for broad generalization. This is very much in the spirit of Unikernels (one of my all-time favorite projects, shoutout to @avsm). The hope here is that one doesn't need to write a declarative spec themselves as agents become more capable.
1
3
2,621
Claude found a vulnerability in the Linux networking stack that allows a KVM guest to panic the entire host by sending just two Ethernet frames over its virtual NIC: lore.kernel.org/netdev/93792…
1
7
1,796