Experiments
What we tried, what actually happened, and what it means for defenders.
RESEARCH
How real attack paths actually emerge.
We research the seams — where software, infrastructure, identity and security controls meet, and where a low-risk finding becomes a real exploit path. We publish experiments, exploit-chain postmortems, benchmarks and control failures, with primary-source citations.
Reproducible research, not “what is X?” filler.
Exploit-chain postmortems
How mediums compose into a critical path — de-identified.
Benchmarks
Harder measures that punish hallucinated exploitability.
Control failures
Where a WAF, IAM rule or endpoint control broke a path — or failed to.
From the blog.
Your Repository Is Executable Input Now
Coding agents changed the repository threat model: configuration, hooks, instructions and tool definitions can influence privileged agent behavior.
Read
MCP Is Becoming a Security Boundary. Treat It Like One.
MCP security is not mainly a prompt-filtering problem. Once an agent can call tools with real credentials, the useful threat model is identity + authorization + untrusted input + execution.
Read
Cloud Context Is Not Enough: Security Testing Needs the Organization's DNA
Real attack paths cross code, identity, infrastructure, controls and organization-specific rules. Product security needs a full-system model.
Read
The AI Pentest Race Is Becoming an Exploit-Proof Race
Autonomous offensive security is attracting serious capital. The durable product is not another alert generator---it is defensible exploit evidence and re-test.
Read