Strong agree on the warm feeling about returning to Jetbrains IDEs, and my wish that their quality could embue the current crop of AI agents with better everything... but I hit too many frustrating disappointments and even a bizarre loss of several hours of work, which is so inconsistent with the thoughtfulness and maturity of their tools I began to suspect aliens...but whatever the source I couldn't keep working with a tool that led to that outcome. Someone let me know when it's safe to try again.
I’ve had similar experiences. Starting to do git checkpoints before each session. And I keep going back to Claude. But I exploring doing a different model - back to monorepos and packages. Having Augment do the last mile so to speak, and it has to write to the APIs I’ve already defined. Also, I’m using my GPT with best next/ts practices loaded in help write a very detailed spec (will share in an article next week) including test requirements. Then have Claude write the code TDD style. Early results are good, but I need to spend more time working with the packages.
Strong agree on the warm feeling about returning to Jetbrains IDEs, and my wish that their quality could embue the current crop of AI agents with better everything... but I hit too many frustrating disappointments and even a bizarre loss of several hours of work, which is so inconsistent with the thoughtfulness and maturity of their tools I began to suspect aliens...but whatever the source I couldn't keep working with a tool that led to that outcome. Someone let me know when it's safe to try again.
I’ve had similar experiences. Starting to do git checkpoints before each session. And I keep going back to Claude. But I exploring doing a different model - back to monorepos and packages. Having Augment do the last mile so to speak, and it has to write to the APIs I’ve already defined. Also, I’m using my GPT with best next/ts practices loaded in help write a very detailed spec (will share in an article next week) including test requirements. Then have Claude write the code TDD style. Early results are good, but I need to spend more time working with the packages.