Comment by thegagne

1 month ago

Speaking of methodical, have you tried AWS Kiro?

It has spec driven development, which in my testing yesterday resulted in a boat load of passing tests but zero useful code.

It first gathers requirements, which are all worded in strange language that somehow don’t capture specific outcomes OR important implementation details.

Then it builds a design file where it comes up with an overly complex architecture, based on the requirements.

Then it comes up with a lengthy set of tasks to accomplish it. It does let you opt out if optional testing, but don’t worry, it still will write a ton of tests.

You click go on each set of tasks, and wait for it to request permissions for odd things like “chmod +x index.ts”.

8 hours and 200+ credits later, you have a monstrosity of Enterprise Grade Fizzbuzz.

Do you think the SDD approach is fundamentally wrong, or that Amazon's implementation was at fault?

  • It sounds like the initial spec is wrong, which compounds over time.

    With SDD the spec should be really well thought about and considered, direct and clear.

    • I think that it has two flaws:

      - It is too machine like in its definition and requirements and misses the spirit of the ask.

      - It very much waterfalls it, without asking for feedback midway, or revisiting the original goals after things have been built. You have such an opportunity to adjust and learn as you go, especially if you keep revisiting your goals and values and re-evaluating your original requirements which may have been flawed.

      Just like with human development, it's rare that your spec is well thought out at the beginning, and impossible that it was comprehensive enough to define a working system.

      I think having goals, vision, and hard requirements make sense, with some guiding principles along the way, but it's very much a journey that requires constant feedback loops and adjustments along the way.

Honestly if you use Traycer for plan + review (I just have it open in different IDE that they support), you can use any editor that has good models and does not throttle the context window.

I am trying to test bunch of these IDEs this month, but I just cant suffer their planning and have to outsource it.