A famous investor let AI write his op-ed. He wasn't sorry.

The story

The op-ed that wasn't hiding what wrote it

Stanley Druckenmiller, the hedge fund manager who once ran billions for George Soros, published a Wall Street Journal op-ed on the bond market, and an AI detection tool flagged it as machine-written before most readers noticed on their own. Critics said the piece proved nobody had really thought it through. Druckenmiller did not apologize. "My name is on the piece," he said. "It's my message."

Why it matters for you: the same week, a developer coined a name for the opposite failure. He called it a "meat proxy" — someone who pastes an AI's answer into a Slack message or an email with no read-through in between, no judgment applied, nothing added. "I don't need a meat proxy in between," he wrote. Same tool, opposite outcome, and the difference has nothing to do with who typed the words.

Brice Challamel, now head of AI strategy at OpenAI, has a name for what separates the two. He calls it the shift from authorship to stewardship: ownership no longer comes from writing every sentence yourself. It comes from how carefully you framed the problem, how hard you pushed back on the draft, and whether you would still sign your name to it. "A good editor doesn't need to have written a manuscript to be accountable for what gets published," he puts it.

Your team is already drawing that line a dozen times a day, mostly without knowing it. This week, ask one direct report the question yourself before you ask them to answer it for you: would you put your name on this one?

The practice

Stop writing the same prompt twice

Most people treat every task like a first-time task, typing out the same paragraph of instructions each time they need the same kind of thing. That is teaching AI nothing. It is feeding it a fish every single day.

Weak, because you write it fresh every time:

Compare this month's numbers to the budget. Flag anything unusual. Tell me what's normal, what to watch, and what to raise with the board. Use the same format as last month. Be specific about how urgent each item is.

Strong, because you teach it once:

I want to create a skill that compares our monthly financials to the approved budget, for deciding what is normal variance, what to watch next month, and what to raise with the board, with every variance ranked by how urgent it is.

That second prompt, said once to Claude or ChatGPT, builds something you can reuse next month, and the one after that, without retyping a word. In Claude, it's under Settings, then Skills — record your screen doing the task once, or hand it that sentence, and it writes the skill for you.

Pick one task you find yourself explaining to a model the same way every week. Give it ten minutes and turn it into a skill this week, so next week you spend those ten minutes on something new.

Higher ground

Stewardship is not a new word for us — it is the older word for what Challamel is describing, and Scripture got there first.

Now it is required that those who have been given a trust must prove faithful.

1 Corinthians 4:2

A steward is not judged on whether their hands touched everything. They are judged on whether what was entrusted to them was handled with care, and whether they can give an honest account of it. That is the test for a tool, too — not who typed it, but whether you would stand behind it.

The next step

If this is the kind of question you want to work through with other leaders instead of alone, Base Camp is a monthly room built for exactly this — AI, faith, and leadership, together. See what a session looks like.

And tell us how this landed. Hit reply with what helped and what you wish we had covered. We read every one.