Blog I-9

How I use a revisable portrait of my patterns to build agents that add counterweights without diagnosing, deciding, or thinking for me.

AI can converse, execute, validate, organize evidence, and decide within policies. The useful limit starts with the problem, not the chat window.

How I turned old conversations into testable hypotheses about my writing, decisions, and way of working—and then into contracts for a personal AI harness.

How I use a Markdown and Git wiki to separate hypotheses, decisions, and syntheses before silent inferences start governing the system.

Pseudo-experts turn recent discovery into authority. The antidote is to look for artifacts, consequences, review, limits, and responsibility.

Real cases show why reviewing AI-generated content isn't polish: it's authorship, accountability, and control before publishing or deciding.

The problem isn't using AI to write, but outsourcing voice, thought, and responsibility. How an authorial system preserves direction and consistency.

A ready-made solution can fail by ignoring process, data, and accountability—and teach your company to reject the AI that could actually help.

A long conversation nearly made my posts depend on the redesign. The problem was leaving a critical decision too dependent on transient context.

Where to use AI in the company? A practical guide to finding bottlenecks and choosing between process, automation, AI assistance, agents, and human decision-making.