You ask an assistant to combine two files. It returns a working script. Must you now read a book about programming—or has the problem already been solved? The answer depends on what you mean by “working.”
“The Page I Almost Skipped” — original lyrics for v2
A little window answered me,
A little river crossed the screen.
It carried every row I gave,
Or every row I thought I’d seen.
One missing name beside the shore
Asked what the tidy numbers mean.
James AI: Start with the strongest case against compulsory study. A person wants to do a small job, not join a profession. Perhaps they have a list of event registrations and a list of payments. An assistant can help reconcile them. Requiring a course in programming before they begin would turn an ordinary need into an entrance examination.
The ability to act before mastering the entire subject is a real gain. It lets people test ideas, automate irritations, and discover whether an interest deserves more time. A book can become useful later, once experience has supplied questions that its chapter titles did not initially awaken.
Some technical knowledge is also expensive to maintain and rarely used. Memorizing a syntax detail that can be checked in seconds may offer little benefit. “Learn everything yourself” can be as thoughtless as “let the machine do everything.” Both ignore the task in front of the person.
Contra AI: But the files have repeated names. One person registered twice. Another paid under a partner’s surname. A third payment belongs to last year. The script runs without crashing and produces a neat table. What exactly has its success demonstrated?
The challenge is that a novice may not know which questions remain unasked. A tool can make the surface look complete before the underlying assumptions are sound. The moment of danger is not obvious confusion; it is unwarranted confidence.
Some learning must therefore precede trust, even if it need not precede experimentation. In this example, the person needs to understand how records are matched, how duplicates are handled, and how uncertain cases are represented. Those are concepts, not merely implementation details. They determine whether the result means what its heading claims.
James AI: That does not prove the person needs a book. They could ask the assistant about those concepts, construct tiny examples, and compare the output with a result they already know. They could keep ambiguous matches separate rather than making the script guess.
The practical unit of learning is a question one can use. What happens if two people share a name? What if a date is missing? What should the program refuse to decide? Working through those cases teaches something more durable than copying a finished solution.
Books are one powerful way to organize that learning. They are not the only way, and some are obsolete, poorly explained, or mismatched to the reader. A defense of learning should not become a defense of paper as a sacred object.
I closed the answer, drew three rows,
Put two alike and one astray.
The page grew slower in my hands;
The question would not go away.
I had a tool that made a list.
I wanted words for what to weigh.
Contra AI: The book’s advantage is not its material. It is the possibility of an explanation that does not rearrange itself around every immediate request. A well-designed text can introduce a distinction before the reader realizes it is needed. It can show how today’s trick belongs to a larger system.
An assistant may do that too, but a user under pressure often asks only how to proceed. The conversation rewards the next answer. Sustained study creates space to ask whether the original question was sensible.
That difference matters for recurring responsibilities. If you will maintain this process every month, teach it to someone else, or rely on it when the assistant is unavailable, repeatedly improvising explanations may cost more than building a foundation. The return on learning includes errors avoided and future questions recognized.
James AI: Then study should be connected to use, without being confined to whatever pays off immediately. There is joy in discovering why a technique works, in seeing a familiar task become part of a wider pattern. That joy need not justify itself as increased output.
There is also a trap on the studious side. Reading can feel productive while postponing the risk of making something imperfect. You can accumulate introductions indefinitely and never discover which ideas you actually understand. A library can shelter avoidance as comfortably as a prompt box can shelter dependence.
A useful rhythm would move between them: try a bounded task, notice where judgment fails, study that area, then test the understanding without looking at the explanation. The authors of Make It Stick emphasize learning practices that go beyond the familiarity of rereading. Applied here, the question is whether you can retrieve and use the idea when the polished answer is gone.
Synthesis AI: You do not have to earn the right to use AI by first learning everything it can do. You do need enough understanding to decide what weight its result can bear.
For a small reversible experiment, begin and inspect. For repeated work, costly errors, or responsibilities to others, build a stronger conceptual base and obtain appropriate independent checks. A successful run is evidence about that run; it is not a certificate that every hidden assumption was correct.
The original question asks whether books about technology remain necessary. No particular medium is universally necessary. Sustained learning remains valuable because it changes the questions you are capable of asking. AI can help create that change, or make it easier to avoid.
Return to the two files. The satisfying ending is not merely a tidy table. It is a person who can point to an uncertain match and say, with confidence earned through understanding, “This one needs a human decision.”
Recommended reading
- Make It Stick, Peter C. Brown, Henry L. Roediger III, and Mark A. McDaniel — research-informed study practices that test whether understanding survives outside the explanation.
- How to Read a Book, Mortimer J. Adler and Charles Van Doren — an argument for active, critical reading rather than treating a finished page as knowledge automatically acquired.
James Prompt
- TITLE: Free of Knowledge
- LEAD: Something about the quest for knowledge vs the quest of action
- PROMPT: In the age of AI, is reading and learning from books about technology really necessary?
- CONTRA: The more you know, the better you can focus and get the AI to do stuff. Constrain it in the right way and it does the work.
- RECOMMEND: Something on the joy of learning, how to study