The Day Anthropic's Own Founder Called for a Pause: A 66-Year-Old's Reflection on Learning Alongside Claude

🎧 Listen to the Discussion(two hosts talk through their thoughts on this article)

This is an AI-generated audio overview of this article, created with Google NotebookLM. Two hosts discuss the article together, sharing their own thoughts and commentary — it is not a reading of the article text.

On 12 September 2026, a major news story broke. The chief executive of Anthropic, a generative AI company, called on the entire industry to deliberately slow the pace of frontier AI development. He went further, warning that if this were handled incorrectly, the entire internet could be taken over by autonomous AI systems within six months to a year. When I heard that rival chief executives were, one after another, expressing their support for this view, I must admit I felt slightly unsteady on my feet. This was because, for the past six months or so, I myself had been using Anthropic's own generative AI, Claude, as a partner for overhauling the technical side of my blog and preparing the groundwork for my investment research. The very people who had built the thing were now saying, "Hold on a moment." In this piece, prompted by that news, I would like to look back on how I, at sixty-six, have journeyed alongside Claude.

The curve of technological progress set against the slower curve of our understanding and institutions catching up, and how wide that gap has grown.

Until I retired at sixty-five, AI had nothing to do with me

For many years, my work involved engaging directly with the people in front of me rather than with data on a desk, so, to be honest, generative AI had long struck me as "a tool for the younger generation". Right up until I retired in 2025, I did only the bare minimum of computer work, and the idea of writing code had never seriously crossed my mind. I felt this way because everything I had done up to that point involved meeting people face to face and making judgements with my hands, my tongue and my nose. I was convinced that things such as the heat of a kitchen flame or a patient's pulse were beyond anything an AI on the other side of a screen could measure.

What I tried, one step at a time

In 2026 I discovered Claude and decided to entrust it with a major overhaul of this blog, adventures-after60.com. At first I did not even know what a "build script" was, and my instructions would often produce results that were wide of the mark. Even so, step by step, I worked through problems such as how to keep a bilingual growing log for the cucumbers in my greenhouse, and how to structure investment research reports so that they read clearly, all the while exchanging messages with Claude. I was able to keep going through this roundabout process because, whenever something went wrong, asking "What happened?" would bring back an explanation that included the reasoning behind it. Once I understood the reasoning, I could see what to instruct next. That repeated cycle was, for me, the very essence of the learning.

The cycle of asking a question, receiving an answer, trying it out, and thinking about the next step based on the result, repeated over and over as I gradually made progress.

The sensation of "losing control" I learnt at medical school

During my five years at medical school, what interested me was not disease itself so much as the process by which the body's internal balance gradually tips, until at some point it can no longer be corrected from within. The situation described in this news report, in which an AI improves itself and continues gaining capability without human involvement, seems to me to overlap in some ways with the behaviour of proliferating cancer cells that I learnt about back then. This is because both share a common trait: the first small changes are difficult to detect, and by the time they are noticed, control has already become difficult. In Buddhist thought there is a concept known as the five aggregates of suffering, the idea that the very act of the five functions making up body and mind becoming too vigorous is itself a form of suffering. Progress in itself ought not to be a bad thing, and yet, if it becomes too vigorous, it may end up causing suffering rather than relieving it. Reading this news, that phrase came back to me unexpectedly.

A difference in tone between Britain and Japan

Living in Britain, I sense that the tone of reporting on AI risk differs somewhat from that in Japan. British coverage tends to focus on the impact on employment and copyright, and warnings about outright loss of control, of the kind raised in this story, seem to have circulated first within expert communities rather than the wider press. Japanese coverage, by contrast, centres on the CEO's own words, the very warning I mentioned at the start, that the internet could be hijacked within six months to a year, conveyed in a way that feels closer to everyday life. Neither approach is more correct than the other; it is simply that the same event can be reported so differently depending on where the focus falls. Living across two countries is precisely what allowed me to notice this.

What the potatoes taught me

This year, I was growing potatoes in my raised bed. Looking only at the leaves above ground tells you nothing about what is happening underground. It is only once you dig them up that you discover they have grown larger than expected, or, just as often, that disease has got to them first. I think what happens inside a generative AI is rather similar. As users, we cannot see directly into the mechanism that produces the answers we receive. That is precisely why it feels significant that the people building these systems are now the ones saying, in effect, that they need time to dig down and check.

A diagram showing the shift from seeing AI as a convenient magic box to seeing it as a tool whose reins still need to be held.

What learning alongside generative AI is giving me in later life

This kind of learning is directly connected to how I keep my mind and body in balance in retirement. Working through something I do not understand, one piece at a time, exercises the mind in much the same way that physical movement exercises the body. It is not free, of course. It takes time, and the effort spent fixing mistakes and going back and forth is not a small cost for someone living on a pension. Even so, I keep at it because there is a reward here that money cannot buy. Reading this news taught me something else: the habit I have built by incorporating light therapy into my daily health management, that of paying attention to the invisible workings within the body, turns out to connect directly to how I think about the invisible mechanism of AI. Even while the light is being applied, I can feel my body gradually easing and relaxing, yet the changes taking place deep within my blood flow and cells remain invisible. It was not only the encounter with the technology itself, but also the discovery of the people who have spoken up about it, that broadened my view in this stage of life.

Why I keep writing anyway

In my fifties, I remember feeling both eager for new things and reluctant about change, at the same time. At sixty-six, that same feeling remains, only in a different shape. Even so, I have no intention of stopping this record. I am not writing in order to receive a response; I simply hope that the record itself, sitting here, might become a small light for someone. Whether AI's advance should be slowed is not something I get to decide. But trying it out, one step at a time, and writing down how I make sense of it, feels like the quiet kind of companionship that is within my reach right now.

Superintelligence: Paths, Dangers, Strategies by Nick Bostrom Kindle Edition Paperback

(As an Amazon Associate, Adventures After 60 earns from qualifying purchases.)