You not knowing your system is technical debt. Agents are capable, but they need steering long-term. You don't need to know every detail, but you should understand at some level how your project works. Agents can do cool things but too many times I've that they just won't unless…
To be nitpicking, that is knowledge debt not technical but I agree. I let my llms write ownership docs for this precise reason. Until I can answer questions about the change I do not approve the PR
I architect the solutions. I agree, I must know the system before doing anything. I use the AI to understand the codebase since I am working on more legacy and non-documented projects. Yes, it does cool things if we steer it correctly.
I have about 2 hours on weekend days. Otherwise it's in the evening after work so rarely get to work on side projects 🙂 And I have so many photos to edit still from past trips 🙈😅
They should, tbh 🤔
But of course, you should not _ask_ for it nor anyone should be forced to implement it. People are free to voice their preferences however 🙂
Yeah really depends I guess on the project. Backing out a bad financial operation might be costly 🙈 Figuring out what tests to run gets complex really fast if you want anything fancier
Yikes yeah.... Little harder in that situation. If you can't just throw more runners at the problem it sounds like you gotta optimize the testing itself. Are you loosing time to something dumb like setup or tear down. I know we had a testing workflow that 2/3 of the time was…
We did that as well. Moved to Blacksmith, optimized many tests, added speed traps, implemented TIA, split logic from IO (so more unit tests), and booting only the parts of framework the tests needs. Manageable so far, but I wonder what else people do 🤔