He Can Name More Things Than He Can Build

I spent a few hours of last weekend doing vibe coding with my 8 year old. By the end of it we had a working website dedicated to 8 to 12 year old kids. One thing for sure, I learned slightly more than him.

He being the inquisitive one, always asked me can he experiment with Codex or Claude Code. I was not sure about it. He came into the session knowing LLM. He knew open source, Github, Gitlab etc. before I could explain him.

The surprise came when we were chatting and he realized I am transcribing it with Granola. After knowing it has a free plan, the next question he asked is if it will sell our voice to anyone. And later on, while discussing, he could explain to me data brokerage in his terms.

At the same time he did not know what a frontend is, what a backend is. He said HTML is four random words put together. He did not know what a repository is but knew Github. I explained him what CSS stands for and till the time of writing this he is not able to recollect it.

This is the shape of new learning. It is like he knows many things at the top of the stack and hardly anything at the bottom.

This is exactly reversed to how we learned software. We started from the bottom, with C as the primary language, and then worked up to something a user could use. He started at the top because there can’t be anything under it at this age. And most of his learning comes from YouTube. Everything he watches has a reason to mention model, token, open source, github. Nothing he consumes explains him basics at the moment.

I can see this inversion pattern in smaller things. He could name ChatGPT models like 5.1, Terra, Luna and could compare them to Fable, Opus, Sonnet and Haiku. He did not know what a haiku is, guessed that sonnet is a small poem and opus is a big one. This is actually ambient learning. The brand names reached him first, then the actual ones. It is like he knows everything at the surface in high definition but skips everything the surface was built on.

None of the definitions I taught him survived the other room. What actually stuck was a bug.

The first version of the site had 3 questions. He played once and got all 3 right. He replayed the quiz and the same 3 questions came back. He was annoyed. I asked him why and he worked out that questions are static. Without much help he asked Codex to build a fix, and the fix was to have a question bank and show random 3 questions. He has not forgotten static and dynamic since then. He forgot definitions of HTML and CSS instantly.

So the rule of teaching is not complicated. Don’t just feed them definitions, let them experience a failure. Explanations handed over after a failure are permanent.

The other discovery was accidental. The best conversation of the day happened when nothing was working. Codex ran 4 to 5 minutes at each stretch, and in those gaps we covered git, open source, chit fund, Subrata Roy, nepotism, data as currency. The planned agenda of vibe coding produced a list. The dead time produced everything.

In the normal coding session there is no idle time. You are always busy. Waiting for a model to respond is free time, and instead of wandering to other windows, it is better to strike a conversation. This is the most valuable part of the conversation.

During this I asked him what he wanted to build next. Oil was out, as oil money is considered bad. Financial services was out, as he did not want to be like Subrata Roy. Using our family’s contacts to open in a second city was what he called nepotism himself, and did not want to be called someone’s son. He asked me about a business that is profitable but not in a Ponzi way.

He has vocabulary for illegitimate money and none for legitimate money. He knows chit fund, what a data brokerage is, what nepotism looks like. He does not know what a margin is. The exact same feed which taught him about LLM also taught him that fast money is usually scam. I do not think it is bad. But it means that the first economic instinct of this generation is suspicion.

Then he made a mistake which I have learnt hard way. His first idea was tools for farmers. I asked him why farmers, why not kids. His answer was simple, wanting to help. He was admiring the problem. With little nudge he was moving to a problem which he himself has.

He wants to start a rent a book business. He got into specifics immediately. Books laid out on his bed, photos, video call, delivery by kids, ten rupees each etc.

Then he climbed up the ladder of scaling. Deliver by kids, one area to other. Suddenly he said nobody will invest in a business so small. The market size argument was discovered by him. I nudged him towards second city, he said he will find people who know there, and the answer was second warehouse.

The sharpest moment came few minutes later. He offered to lend his Kindle. I asked him why. He had no answer. He was never taught competitive moats. He felt one.

Somewhere in the middle of all this I asked him who gets the credit for all this, you or Codex. He answered very clearly that he wrote the prompt, the spec, he should own it. Adults have been arguing about it for years but he settled it before he could spell repository.

I want to be honest about what this article is evidence about.

It is not evident that my son is unusual. It is evident that proximity is unusual. He sits near a CTO, a functional pipeline, a brand being built. Attends whiteboarding sessions he was not invited to. Any child having this exposure will sound like that. The interesting thing here is the access. It is not about him. Probably he has figured it out and doesn’t want to use family contacts.

And this exposure has a cost attached to it. He can name more things than he can build. The gap between vocabulary and comprehension is what consumption produces. Left alone, the gap hardens into adult version. Where people talk about AI fluently and never ship one.

This is the whole reason for doing it at eight. Not to make him a programmer. To make sure that the thing which he named becomes a thing that existed. So the distance between two never gets comfortable.