one thing i had a ton of issues with in my code quality was not using modern python typing preferences - list, dict over List, Dict in 3.9+. the root of that is because claude code is just trained on a bunch of old code, and using claude code extensively, i got flamed in my review. which was very helpful because now i know better. i love making mistakes, even though it comes at the expense of my pride—that's something i should let go of more often.
also learned about the benefits of asyncio over threadpool in python. essentially it uses a lot less memory because there's only one thread managing the event loop, versus threadpool which spawns multiple threads.
ASYNCIO (Single Thread)
[Event Loop]
│
┌─────────────────┼─────────────────┐
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌─────────┐
│ Task 1 │ │ Task 2 │ │ Task 3 │
│ [API1] │ │ [API2] │ │ [API3] │ ← all tasks
│ waiting │ │ waiting │ │ waiting │ started
└─────────┘ └─────────┘ └─────────┘ instantly
Memory: ~1 thread, minimal overhead THREADPOOL (Multiple Threads)
┌──────────┬──────────┬──────────┬──────────┐
│ Thread 1 │ Thread 2 │ Thread 3 │ Thread 4 │
├──────────┼──────────┼──────────┼──────────┤
│ [API1] │ [API2] │ [API3] │ [API4] │ ← 4 tasks
│ waiting │ waiting │ waiting │ waiting │ running
│ │ │ │ │
│ [API5] │ [API6] │ idle │ idle │ ← tasks 5-6
│ waiting │ waiting │ │ │ must wait
└──────────┴──────────┴──────────┴──────────┘
Memory: ~4 threads, higher overhead per thread