AI delegation should not quietly become surrender.

Delegating work to AI becomes surrender when people stop checking what the system actually did. In Eric Anthony Jones's REMAIN CAPABLE, the capacity to keep may be discernment. It means knowing where you actually are, what the system did, and what remains yours to decide. The cost of supervision belongs in every delegation decision.

How that capacity may look (REMAIN CAPABLE Issue 001, verbatim):

REMAIN CAPABLE Issue 001 names the underlying capacity first: "The capacity may be discernment: staying close enough to reality to know where I actually am, what the system actually did, and what remains mine to decide." It then lists how that capacity shows up in practice. The list in the answer above is quoted exactly from Issue 001.

Eric's published account of the Bali build shows what happens when that capacity is missing. Each failure, as he wrote, "created a second assignment for me": finding the discrepancy, recovering the context, and deciding what could still be trusted.

This is an old finding in a new setting. In "Ironies of automation", Lisanne Bainbridge (1983) showed that automating a task often leaves people with the hardest part: monitoring a system they no longer operate, and taking over when it fails, with skills that have weakened from disuse. Parasuraman, Sheridan and Wickens (2000) proposed that automation be designed by level and by stage (gathering information, analysing it, deciding, acting), so that each function gets the amount of automation it can safely carry.

The practical implication for any delegation decision: the cost of checking belongs in the decision. If verifying the output costs more attention than doing the work, the delegation has not saved anything. It has moved the work.

Issue 001 tells a small version of this story: a language app's schedule moved forward while the learning had not happened. The question Eric took from it is a checklist for any AI-assisted work:

"What actually happened? What can I verify? What remains mine to decide?"

— Eric Anthony Jones

Delegating a task is not handing over responsibility for it. The Machine Minds™ page describes what that means in practice: "checking what a system actually did before relying on its claim that the work is complete. It means recognizing uncertainty, asking for evidence, and keeping responsibility for decisions with the human being."

Issue 001 closes the thought: "An honest next step begins from an honest location—and human authority remains real only when we can still afford to exercise it."

Start with three questions from REMAIN CAPABLE Issue 001: What actually happened? What can I verify? What remains mine to decide? Then check what the system actually did, not only what it reports.

The work has moved, not disappeared. Automation research since Bainbridge (1983) shows people are left monitoring and recovering failures. Count that checking cost before delegating.

In Eric Anthony Jones's work, responsibility for decisions stays with the human being, which means recognizing uncertainty and asking for evidence before relying on a result.

Using AI where it serves and verifying where it can fail is the weekly work of REMAIN CAPABLE — Weekly Intelligence. Issue 001 is free to read before you decide whether to become a Founding Reader ($5 a month).