Julian Neagu

Middleware order can save you a lot of debugging time. I’ve learned to keep auth and input checks clear and early, especially when building small APIs.

I've found that once you run a few agents, the hard part isn't getting answers. It's knowing what happened between them. Seeing the hand-offs makes debugging feel a lot less like guesswork.

I like the idea of getting the mockup approved before letting AI touch the code. Fixing the direction early saves a lot of back and forth.

I’d also keep an eye on rate limits before pushing concurrency too high. It’s easy to make things faster on your side and just move the bottleneck to the model API.

I’ve found that giving an agent too much access is easy to overlook when you’re trying to ship fast. Keeping it locked to the project folder feels like a simple win.

The cost per successful task is the part that matters. A cheap model that needs two retries can get expensive fast.

The point about agents installing packages without review hits close to home. When you're building fast, it's easy to trust whatever the tool suggests. One bad dependency can turn a small shortcut into a big mess.

The “build it by hand first” part is spot on. I’ve found the same thing with agents — frameworks make sense much faster once you’ve fought with the loop yourself.

The cents point is underrated. I’ve had AI outputs behave way better once the schema makes units explicit. totalAmountCents leaves a lot less room for weird guesses.

The repeat runs are the part that stood out to me. A model can look solid on one pass, then miss the same kind of task later.