Articles
Long-form writing on tech, culture, and the edges of the internet.
Cognizant's bench is shrinking by design
Cognizant's automation push isn't a productivity story - it's the collapse of the services pyramid. What's actually changing, and why most firms will get the transition wrong.
Copy.fail has been root since 2017
Copy.fail turns an unprivileged Linux user into root via a copy_file_range credential cache flaw. Reachable since 2017. Telemetry gaps explained.
Encrypted files are writing back to disk
Active ransomware event analysis from an operator perspective: what failed, the underlying mechanism, and the conditions that must now hold.
OpenAI's security plan protects nothing yet
M. Hale on the OpenAI cybersecurity action plan: provider-stated intent is not a control, and the consumer still owns the boundary.
CVE-2026-3854 puts GitHub inside your trust boundary
CVE-2026-3854 enables RCE on GitHub.com and Enterprise Server. Why platform compromise becomes customer compromise across identity, secrets, and artefacts.
Managed Agents pricing is an architecture decision
Claude Managed Agents pricing isn't a cost center - it's an orchestration lever. Here's how to evaluate it against real total cost of ownership.
ShinyHunters exfiltrated Cisco source through Trivy
ShinyHunters exfiltrated Cisco source code through Trivy. The scanner inherited the runtime's identity. The runtime held everything.
Your second factor is a phone call
SMS 2FA on PayPal is a routing decision, not a credential. The session cookie is the boundary, and attackers have already routed around the factor.
A postcard breached a warship
A 5 dollar Bluetooth tracker hidden in a postcard broadcast a 585 million dollar warship's position for 24 hours. The control that failed was classification.
Binding 65535 ports is the easy part
Architecture and evasion realities of an LLM honeypot binding all 65535 ports - TPROXY, latency tiers, fingerprint defence, and detection traps.
CISA flagged a 17-year-old Excel flaw
A 17 year old Excel flaw is being actively exploited and flagged by US cyber defence. Operator analysis of what failed, why, and what must change.
Engineering teams keep granting agents production database writes
AI agent vulnerabilities are systems engineering failures, not security failures. The fix is architectural containment, not better prompts or guardrails.