ONE OF THE REASONS I'M FINDING LLMS SO MUCH MORE USEFUL FOR CODING THAN A LOT OF PEOPLE

Summary from the TLDR newsletter

All of Simon Willison's code has automated tests. This massively de-risks his use of large language models. He uses AI to write code and then proves that code works with tests, then uses AI to refactor the code until it looks good, keeping the tests green the whole time. Programmers who find the technology to be more of a hindrance than a help in their work may need to check the health of their test suites.