Nothing Is Making Me

Apparently I must now make time to think. I had been under the impression that thinking was what kept interrupting the things I was trying to do. Now it needs protecting from the things that do them for me.

This is going to require a better man. I resent that. One of the attractions of machinery was that it worked for the man I already was.

In Business Insider on November 5, Henry Chandonnet reports on a warning from Addy Osmani, an engineer on Anthropic’s Claude Code team. Osmani writes that, increasingly, nothing in his day forces critical thinking on him because so much of his work goes through agents. His answer is to build that thinking deliberately into the day.

Not stop using the machines. He says he offloads work all day and recommends it. He is quite happy for people to delegate entirely where the stakes are low. This isn’t a demand to suffer over every small task. It is advice from a man helping to build the thing which has made the advice necessary in his own working life.

I find his observation more disturbing than a warning about a wrong answer.

A wrong answer gives me a job. Find the mistake. Correct it. Complain about whoever supplied it. I have qualifications in the third part. But suppose the answer is right. The work gets done. Whatever used to make me stop, guess, fail, reconsider and eventually find a way through has ceased to be necessary to finishing it.

Now I have to remember to do some of that anyway.

I have never been good at doing things anyway. I need an unreasonable desire or an inconvenience with consequences. A sentence won’t come right. Somebody has misunderstood me, possibly with excellent reason. A bill has arrived which I cannot pay by appreciating it. My mind gets involved because leaving it out has become less comfortable than using it.

None of this proves that the old arrangements were good. A great deal of work can disappear without taking anything worth keeping. I don’t need to preserve my ability to be furious at a badly designed form. I could lose that entire accomplishment and remain a difficult man. Other people have things to finish. I have things to finish. I am not going to ask a useful machine to become worse so I can feel employed between two sentences.

But I didn’t acquire everything useful by deciding to acquire it. Sometimes I wanted the thing at the other end and discovered, with considerable annoyance, what I would have to learn to get there. It didn’t require me to believe in the long-term preservation of Henry. I was trying to finish a paragraph.

Osmani has a question on his list: “What am I no longer practicing?”

There is an unpleasant amount of choosing hidden in that. Not just what I will let a machine do. What kind of man I am prepared to stop being able to be without it. I could answer grandly: keep the judgment, delegate the drudgery. A magnificent division. Everything flattering about my contribution stays mine. Everything I am tired of can leave.

Then I have to identify the parts.

Take the bad first attempt. Often it deserves to die. There is no reason to build a museum around every awkward sentence I have written. Yet a bad attempt sometimes tells me what I am attempting. I begin wanting to say one thing and find, through making a mess of it, that I meant another. The finished version could be better than anything I would produce. It could also arrive before I have noticed that I want something else.

This is close to the distinction Osmani makes, in Chandonnet’s account, between offloading and deferral: the machine’s output becomes the next input. There is now something to respond to. I can disagree with it, sharpen it, ask for another version. I am occupied. I may be thinking hard. But I have started with the thing it made available for me to think about.

His question “Did I formulate a hypothesis before asking?” lands harder than a lecture about laziness. Before asking. That is the troublesome part of the schedule. Once there is an answer, I have plenty of opinions about it. I can spend an hour improving something without having spent a minute deciding whether it was the thing I wanted to begin.

I don’t take this as proof that agents are damaging everybody’s brain. Chandonnet explicitly says the evidence doesn’t yet settle whether AI is making us less capable or changing how we think. Osmani’s account is a warning about his working day, not a diagnosis of mine.

Nor does he leave the practical work unguarded. He recommends making verification a habit, including asking the agent what it could check and what it couldn’t. Good. I want the work to be right. I also want to be able to do something besides approve work which has arrived right. Those wishes can agree without being the same wish.

The paragraph can be delivered on time without any of this troubling the person who asked for it. He wanted the paragraph. It isn’t his business whether I had a first thought before requesting one. I could satisfy him every Friday and keep putting off the part that concerns me. Nothing would arrive marked late.

I have been putting off organising my life for years. Apparently my mind has been added to the job.

Source: An Anthropic engineer warns about ‘cognitive surrender’ as AI agents take over his work (Henry Chandonnet, November 5, 2026; Osmani’s remarks as reported by Chandonnet.)