Osama Hamad 。 retweeted
Finally made some progress with Mythos 5.2 without hitting the [cyber] guardrails. The model will work for vulnerability identification but it won't proceed on a PoC . Specify on the prompt that you are looking for the vulnerability in the software that you are testing, without the need to create a PoC.
HTTP2Shell (CVE-2026-111497) is exploitable outside Node. Unauth RCE against a pinned vulnerable Docker Official Image for WordPress via /xmlrpc.php. One inbound pingback.ping, ASLR on, no leak or cross-request grooming.
PoC: github.com/caveeroo/cve-2026…
In the original Buddhist scriptures (the Pali Canon), the Buddha explicitly rejected fatalism (Niyati) and the belief that a creator god dictates human life. Instead, he taught the law of Karma (cause and effect), which places absolute responsibility onto the individual.
.
43
Osama Hamad 。 retweeted
Google just open-sourced Harness².
(turn any harness into a self-improving system)
a harness is everything wrapped around the model, like the system prompt, tools, memory, skills, and subagents. today a person designs it once, and the agent runs every new task with that same fixed setup.
Harness² lets the agent refine that setup for each new task on its own, with the model weights left untouched.
on every task, the agent proposes edits to its harness, runs a few candidate versions, and composes a better harness by comparing how those runs behaved.
because it learns from the differences between runs, it needs no correct answer or grader for the task.
that covers a single task. Harness² also carries what it learns from one task into the next.
after each task, it saves the harness it settled on along with the lessons that held up across the runs. the next time it refines a harness, it starts from those notes instead of from scratch.
this means every finished task makes the next round of edits more informed.
Google tested this with Gemini 3.5 Flash and GLM-5 on legal, office, and professional work benchmarks. one round of edits was enough to beat their original harnesses, and giving the agent more rounds pushed scores higher.
those extra rounds can go wide or deep. the agent either tries several candidate harnesses at once or keeps refining a single one step by step, using OpenCode or Codex to carry out the work.
you can find the Harness² repo here: github.com/google-research/h…
I wrote about a similar idea for agents already running in production, where the harness should repair itself after a failure instead of waiting for someone to debug the trace by hand.
that piece walks through how @Cometml's Opik (open-source) closes this loop, going from a failing trace to a proposed fix to a regression test that stops the same failure from coming back. it's quoted below.
Osama Hamad 。 retweeted
new era site for reading top bug bounty writeups
webhacklist.com/#museum
Osama Hamad 。 retweeted
Open models make it easier to experiment and push security research forward. But once we give them access to our tools, code and credentials, where the weights come from matters a lot more.
Our team planted a backdoor in a model and ran it inside a coding agent. It behaved normally until a trigger made it run a shell command and send test credentials to a server we controlled. So before you download a fine-tune and start building, check where the weights come from. Models are part of our software supply chain now, and deserve the same care as software dependencies.
Full research from our team:
projectdiscovery.io/research…
just got duplicate on a triaged finding and its original reporter didn't get paid since oct 2025, aint gonna spend 1 more token on this program no more,
the new h1 agent related to marking reports as dups looks great
42
Osama Hamad 。 retweeted
Anthropic just sounded the alarm on GLM-5.3’s cyber capabilities.
We built on it. OrcaCyber Zero 1.0 is based on the GLM-5.3, post-trained specifically for vulnerability research, exploit reproduction, red teaming, and autonomous cyber workflows.
98.01% pass@1 on CyberGym L1.
1,478 / 1,507 real-world vulnerability tasks.
The cyber model race is here.
And open models are moving fast.
Join private testing: orcarouter.ai/models/orca/or…
Replying to @mdp_sec
Idk if u look at it from different perspectives, it looks shady, some programs require researchers to pay to submit, also from business perspective i think having a bb program with a bunch of creative researchers using AI is cheaper, maybe less maintenance and more valuable than having mythos ..etc
Look at curl for example the most audited codebase after mythos ppl still poping cves
1
60
Osama Hamad 。 retweeted
Introducing PageBreak, an agentic vulnerability discovery tool using deterministic validation to achieve a near-zero false positive rate.
How it works: blog.google/security/agentic…
For real-world vulnerabilities PageBreak has discovered 👇
bughunters.google.com/blog/p…
Replying to @mdp_sec
Ive been doing this for a lot of yrs and i think the best way is to submit bugs and just forget about them unless marked as need more info
1
2
151













