We can create some tests, but also focus a lot on ingesting and understanding existing testing (including migrating it over).
As for defining "tech debt" we are not really in that business. Think of Grit like a very patient junior developer: able to do lots of work, but not necessarily who you want defining the application architecture. We don't presume to know your codebase better than you.
Instead, our sweet spot is things like:
- Upgrading a critical business application from Angular 4 to Angular 15 (or even converting it React)
- Taking an existing test/mocking approach used for 10% of database models and inserting equivalent tests on other models
- Automatically removing dead code after feature flags are rolled out broadly
Nice. I’m very excited to see where the “jr dev ai” goes in the next 5-10yr.
I am a little worried more broadly about jr devs not getting work and thus never getting the experience to become sr devs. I imagine jr devs will come from locations where an ai is too expensive.
I think the tasks that AI will take first are broadly the tasks where jr devs would not be gaining much useful experience anyway. For example, the rote work involved in upgrading an application from Angular/React X to X+1. The available AI tools are still very bad at novel tasks, architecture/design, etc.
I hope that it can create tests for the code that changes to ensure that it retains the same outputs (including side effects) for the same inputs.
But I have to ask: how do you define tech debt? What about these scenarios:
1) comments above the code say “do not touch, here be dragons” ;)
2) code change velocity is currently zero, team used to care and wanted to change to something modern, no one cares anymore.
3) above but with good testing
4) active code velocity at moderate level with ok tests but old/bad patterns, and is an interchange for various parts of the system
5) high velocity code change, no tests, known bug utopia, zillion a/b tests, marketing wants changes hourly. 8)