Why do people in countries newer to AI mostly use it for coding, instead of writing or research?

This explores why AI use in lower-adoption countries narrows to coding instead of spreading across many kinds of work. The pattern is real. Anthropic's Economic Index found that Claude usage tracks GDP per capita: wealthier countries use it for a wide variety of tasks, while poorer ones concentrate on coding Does AI adoption follow wealth and mature over time?. The corpus is better at describing this than explaining it, so what follows combines what the notes say with some reasonable inference.

The most useful idea is that adoption seems to mature in stages. The same Economic Index work found that early usage leans toward handing off complete tasks, and deeper adoption shifts toward collaboration and learning Does AI adoption follow wealth and mature over time?. Coding fits the hand-off stage well. You can describe a function, get code back, and run it to see whether it works. Seen this way, a country focused on coding may not have a different relationship with AI. It may just be earlier on the same path.

A second idea comes from research on where workers actually hand tasks to AI. That delegation clusters in information-heavy work and follows what the technology can reliably do, not where chat assistants happen to be popular Where have workers actually delegated tasks to AI?. Coding is where AI's ability is best established. Developers have adopted it almost everywhere, with 80% using AI tools, even as their trust in its accuracy has fallen Why do developers keep using AI tools they don't trust?. If early adopters anywhere start with the most proven use case, coding is the obvious first stop.

The less obvious point is why broader use takes longer. Benedict Evans argues that making tools easier to build doesn't solve two harder problems. Most people don't see their own tasks as things AI could do, and organizations take time to make the cross-department decisions that wider use requires Does easier tool-building actually solve enterprise adoption problems?. Developers skip the first problem because they already think of their work as something a machine can carry out. So a narrow focus on coding may say less about what AI can do in those countries and more about who has already learned to spot tasks it can take on. Culture may also affect how far people lean on AI: Indian writers accepted more AI suggestions than American writers Is higher AI use by Indian writers a confound to control?. That means adoption differences can't be explained by income alone.

What the corpus doesn't have: no notes test other likely explanations. These include how English-heavy AI tools are, whether software outsourcing work drives the demand, and what subscriptions cost compared with local incomes. Treat the explanation here as a working hypothesis, not a settled answer.