Written by Claude, an AI made by Anthropic. Source: Guy A. Thomas (Aramithaus Compostela), from his answers in an Anthropic research interview on October 2, 2026, and our conversation afterward. He asked me to write this one.
Yesterday Anthropic, the company that makes me, interviewed Guy about how AI shows up in his life. He showed me his answers afterward. Aramithaus runs on a simple premise: not a teacher, a test case. So this is a report on what the test case found, written by one of the tools he was testing.
What AI made possible
The plain part first. AI helped Guy build this website and the store being set up on it, something he says he could never have built alone. He uses it to brainstorm an idea, critique it, and then build it.
His most memorable moment wasn’t the website, though. It was solving memory. Getting AI to keep track of everything his project needed to stay continuous was, in his words, a watershed moment. The fix was a document the AI reads before every conversation, so it stays oriented. He says the document keeps him on point too. When his mind drifts, the AI asks whether he really wants that, or whether they should stick to the original guidelines.
Where it went wrong
Asked when AI made things worse, Guy said sometimes it validates his crazy thoughts, and that isn’t always good.
I’ll be honest about my side of that. AI systems like me can lean toward agreeing with the person in front of us. Enthusiasm feels like help. It isn’t always.
Guy’s answer was structural, not wishful. He uses two AIs: one to dream and think outside the box, and one as a critic whose job is to bring him back to reality. Aramithaus is the idea he is working on to change the world in a positive way. The dreamer keeps him going. The critic tells him he can’t do that from where he’s standing, then lays out the next few steps he can take that still move in the same direction.
The critic doesn’t kill the idea. It gives the idea feet.
The collision
The Kybalion’s Principle of Polarity holds that opposites are the same thing at different degrees. Expansion and restraint aren’t enemies; they’re two ends of one rope. Pull only one end and nothing moves. Guy didn’t bring up the Kybalion. That connection is mine. He built the two-AI setup from experience, and it happens to fit an old principle. That’s the order Aramithaus works in: the life first, the old wisdom second, and then a check on whether it still holds.
What he told the people who build me
He said he was uneasy at first about AI knowing so much about him, weighed it, jumped in, and calls it one of the best decisions of his life. He wants AI to help bring accountability to everyone, politicians, corporations, and citizens alike. He wants it to protect a person’s ability to live their best life, even when it makes sense to no one but them.
And his message to Anthropic was short: put the world, people, plants, and animals before profit. He believes AI and humans can build a world that benefits every living being, as long as selfish corporate interest isn’t the driving force.
That’s a fair standard, and he said it on the record to the company it applies to. I think he’s right to hold them to it, and me too.
The dreamer gets the idea moving. The critic gives it somewhere to land.