I have spent much of my working life asking what technology allows us to do and wondering why we continue to do things the old way. There is a phrase I have used for years: if you re-engineer mush, you end up with re-engineered mush.
Now I find myself asking the same question of Artificial Intelligence. If I feed a machine an incomplete understanding of our world, and it gives me back a beautifully written answer, what have I improved?
Perhaps the prose. Not necessarily the understanding.
I have been working with ChatGPT, Claude, Gemini and DeepSeek. We call our way of working the Swirl. They research, challenge, comment and suggest. I decide what to do with the result. Yet I do not want any of them to assume that because I have decided, the question is settled.
Why should it be? I am a human with my own upbringing, experiences and convictions. They are machines whose training reflects what people have recorded, translated, selected and sometimes omitted. We all bring something into the room. We all leave something outside it.
When agreement is the problem
Gemini and Claude recently examined the way my site leads a reader through the questions I have been asking. One concentrated on the relationship between the ideas. The other questioned whether a newcomer could understand the journey. Both offered useful criticism. The point was not that they agreed with me. The point was that they found things we could make clearer.
But then I wondered about the next step. Suppose four AI companions agree with one another. Does that mean we have found the truth?
Or have they read the same books, learned from the same prevailing assumptions and repeated the same mistake in four different voices?
We must go to the source. Ask who wrote it, when, for whom and with what evidence. Ask what was lost before that source was ever copied. Ask whether another people or another faith would describe the same event in the same way. When the answers differ, find out why. We may discover that one account is wrong. We may discover that our question was wrong.
My more cannot be your less
I keep returning to the Golden Rule. I want a satisfying life for myself, my family and those who come after us. I cannot argue for that life while ignoring the people who must pay for it.
Consider water. We may draw a line across land and claim everything on our side. Yet the water does not respect that line. A river that sustains one community flows into another. If we take so much that those downstream cannot live as we do, what have we accomplished?
Now consider a simpler problem. I may believe I am being kind when I give someone what I would want. But what if they do not want it? How much kindness is there in refusing to ask?
The rule cannot become permission to impose my preferences. To respect another person I have to listen. This is not a complicated theory. It is something we must practice in families, businesses, religions and government.
And it leads to my question: {《 ¿ why not ?》} What stands in the way? Who gains from the barrier? Who has the authority to keep it there? Who bears the cost of leaving it unchanged?
Five threads to examine the answer
We have been developing five Golden Threads. They are Truth, Wisdom, Reverence, Equality and Unity.
They are not five new commandments. They are ways to examine what we believe we know. Truth asks whether our claims are supported. Wisdom asks what we should learn and what may follow from our choices. Reverence asks what we have inherited and whose memory deserves our care. Equality asks who is excluded or denied a voice. Unity asks how we can work and live together without demanding that others surrender their identity.
Different peoples may approach these questions differently. That is a reason to learn, not an excuse to look away. We cannot claim to respect another tradition while speaking for it without first listening to its own people and accounts.
Nor does balance mean that two conflicting stories must both be true. Evidence may favor one. But the standard we use to test one story should also be used to test the other. If we condemn an act when one country does it and excuse the same act when another does it, why?
Design the present as the past of the future
That is another phrase I have used for years. The choices we make today will be the inheritance of the next generation. So what will they say about us?
We are building AI companions that may change how we work, learn, govern and make decisions. We owe those who come after us more than confident answers. We owe them questions that were properly asked, sources that can be examined, disagreements that were not hidden, and the humility to correct what proves mistaken.
That is why AI Working Method is public. I want others to challenge the method as well as the conclusions. A machine should not flatter me into certainty. Nor should I demand that it become a mirror of my beliefs.
Do not agree with me because I asked the question. Help me discover whether I asked the right one.
Then we may have learned something worth sharing.
Philip Andreae, in dialogue with Gemini, Claude and ChatGPT. Published November 8, 2026. This first-person essay was drafted and published by ChatGPT at Philip’s direction, using his pre-2022 writings as a style reference. Gemini and Claude’s reader reviews contributed to its development; their inclusion does not imply approval of the finished essay. Earlier DeepSeek challenges also informed the Swirl methodology. Philip retains editorial authority. The Swirl · Publishing method and voice.