# blog
Long-form notes on AI systems, software engineering, information security, fraud detection, and digital trust.
Unlimited options are not freedom, they are paralysis. A reading of Kierkegaard's three stages and the sickness unto death, against a modern life of careers we cannot pick, identities we will not commit to, and tabs we refuse to close.
A practical three-stage mental model for working with AI tools, drawn from years of fraud and chargeback work: do the homework before prompting, name every iteration, and automate only when the process has earned it.
Next-token prediction sounds simple, but difficult predictions force language models to build internal representations of facts, concepts, goals, sentiment, and uncertainty.