Technical Writing
Tech Blog
Cloud security, identity, infrastructure, and research. Deep dives, tools, and practical guidance.
Synthetic Authority and Constrained Probabilism: What Overloaded Minds Teach Us About AI
From Milgram to cognitive overload to modern agent systems, none of this is new. We are rediscovering decades of psychology the hard way - at machine speed.
MCP, Latency, and Constrained Probabilism
MCP doesn’t fail because of network latency. It struggles because of semantic latency, context rot, retries, and probabilistic drift. This post explores why, when MCP fits, when it doesn’t, and why Recursive Language Models may change the equation.
Capability ≠ Obligation: When Agentic Systems Start Hiring Humans
Agentic AI systems that can directly commission humans for real‑world tasks represent a qualitative shift in risk. This is not about hype or Hollywood. It is about incentives, accountability, abuse, and physical safety.
From Clawdbot to GAINet: When Agent Experiments Outrun Accountability
From Clawdbot to GAINet: When Agent Experiments Outrun Accountability This post didn’t start as a philosophical musing about AI. It started with a very practical, very familiar security reaction: “Why on earth is this …
Allowing ARM for Dev Box: When Portals Impersonate Users (and How to Avoid It)
Microsoft recommends allowing the Azure Service Management API from BYOD to enable Dev Box access. That guidance quietly expands the blast radius far beyond Dev Box itself. This post walks through why that happens, what’s really going on at runtime, and an undocumented deep-link that avoids the problem entirely.
OID-See v1.0.1: Small Release, Sharper Edges
OID-See v1.0.1 tightens scoring logic, improves accuracy around app role assignments, ownership, and deception signals, and lays the groundwork for more trustworthy Entra ID app risk analysis.