In 2028, the question will be who spent more money on lawsuits, and who spent more money on consultants to clean up their code base.

Jokes aside, code tools are best used in the hands of someone who is already trained and can verify bad code, and bad patterns at a glance.

AI code passes many tests. So does a lot of code written by us, for ourselves. When the code gets in front of users, especially the kind of genius users who learn how to fly by forgetting how to fall, then we learn many good habits.

In 2027 we'll have LLMs downloaded to our devices that are as good as Claude Code is today. (But as I have seen, as the leading edge of this stuff is always cooler than what you can run locally, we'll not be satisfied then with today's Claude Code.)