Expert analysis on cybersecurity, artificial intelligence, encryption, networking, and emerging technology — bridging the gap between complex systems and practical understanding.
In April I ran a multi-agent FedRAMP investigation on Cloudflare Workers primitives. One agent flagged an unresolved question about AWS Bedrock GovCloud authorization. Checking it changed the conclusion — and matters if you are planning a Workers AI remediation project.
45 captures. 5 routing policies. 3 traffic matrices. A 7-node containerlab topology running on a single server. The experiment was designed to test whether Hodge decomposition could separate two sources of network flow structure. It could. But not the two I expected.
A new security primitive for AI agent systems. Instead of hiding secrets behind locks, build systems where secrets only materialize inside protected execution boundaries.
Part 8 of the PPOS series. Proofs tell you the system should work. Monte Carlo simulation tells you it does work. Under load, under failure, and under adversarial conditions.
Part 6 of the PPOS series. When your production system crashes mid-transaction, what happens to your data? Event sourcing and exactly-once semantics give you a deterministic answer.
Part 5 of the PPOS series. When you frame production operations as a discrete event system under supervisory control, you get formal safety guarantees that no amount of process documentation can provide.
Part 4 of the PPOS series. Optimal production batching reduces to bin packing, which is NP-hard. Here's what that means practically and how greedy heuristics get you close enough.
Part 3 of the PPOS series. Nine invariants that must hold at all times, proven to survive concurrent operations. This is what separates a workflow from a specification.
Part 5 of the PDKS series. In any niche, multiple domains compete for authority. Replicator dynamics from evolutionary game theory reveal when equilibria exist, and when winner-take-all is inevitable.
Part 2 of the PPOS series. Every production workflow is secretly a finite state machine. Making it explicit eliminates the class of errors caused by 'I thought the rule was different.'
You can export your ChatGPT conversations. You just can't read them. Here's why OpenAI's data export is technically compliant but practically useless, and what I did about it.
Part 4 of the PDKS series. When competitors, bots, or algorithm changes try to destabilize your authority, how much damage can they actually do? The math gives a bound.
Part 2 of the PDKS series. Authority convergence isn't a hope. It's a provable property. Here's the contraction mapping that guarantees it, and what breaks the guarantee.
Part 1 of the PDKS series. In AI-mediated information economies, authority doesn't just happen. It converges, or fails to. Based on structural properties you can formally specify.
AI can write code. Someone has to govern it. A framework for making AI-driven software changes safe, auditable, and usable at scale.
Part 3 of the Nemean Lion World Fabric series. When your system models the world perpetually, raw data storage grows without bound. Structural abstraction keeps it sublinear.
Part 3 of the Adaptive Enterprise series. Digital twin technology has transformed aerospace and manufacturing. Here's what happens when you apply it to business operations.
Part 2 of the SPERA series. Every change to a system impacts other systems. Most organizations track this through memory and meetings. Here's how to track it through architecture.
Part 2 of the Nemean Lion World Fabric series. When you aggregate models from untrusted nodes, you need math, not faith. To guarantee convergence.
Part 2 of the Adaptive Enterprise series. The science of organizational adaptability was formalized fifty years ago. Most software still ignores it.
Part 2 of the SPC series. Classical web systems are Markovian. They react only to current state. What happens when you build systems that remember how users arrived?
Part 1 of the Nemean Lion World Fabric series. Most digital twins predict what will happen. Causal digital twins explain why, and that difference changes everything.
The traditional castle-and-moat approach to network security is dead. Here's why zero trust isn't optional anymore and how to start implementing it.