Writing
Showing 1–12 of 90 articles.
The Floor, the Ceiling, and the Template
Templates make work scalable by reducing variance: they raise the floor, lower the ceiling, and force us to decide what must remain distinct.
What an Apology Is
An apology is the acknowledgment of a debt — duty, breach, harm. Why some apologies should be refused, and why forgiveness is what makes the debt payable.
Conditional Acceptance
A blanket no insults; an unconditional yes breaks the economics. Attach one reasonable condition — and see what the requests were actually worth.
Governance Is a Record, Not a Judge
A governance system should keep the record and flag discrepancies — never judge. Why resolutions bind dissenters, and when acting without authority can still be right.
Legitimacy
Legitimacy isn't whether a decision was reasonable — it's whether it was reached the agreed way. Why procedure pursues finality over truth, and when it must evolve.
The Standard
Standards make iteration and scale possible by exposing defects, enabling cure, naming the judge, and calibrating how high the bar should be.
The Judge
The essential human position in the AI era is the judge: direct the inquiry, develop the record, and remain answerable. Judgment runs on attention, capacity, and memory.
Trust Is the Wrong Question
Trust is the wrong framework for AI agents: humans must define mandates, permissions, boundaries, backups, and accountability for every consequential run.
Trace the Downline
AI systems should be judged by tracing their consequences to human footing: whether they strengthen subsistence, health, capability, and flourishing—or degrade them.
Selfish Respect
Respect for others cannot subsist without respect for self. Self-respect begins with knowing your values, honoring them, and maintaining the self you present.
Response-Ability
Responsible AI at scale depends on preserving human response-ability through bounded mandates, checkpoints, reversible failures, records, risk, and respect.
An Ecosystem of the Accountable
Why stable AI ecosystems require more than distributed value: every agent must act under a named human principal who remains answerable for the outcome.