I think one fatal flaw is letting the agent build the app from scratch. I've had huge success with agents, but only on existing apps that were architected by humans and have established conventions and guardrails. Agents are really bad at architecture, but quite good at following suit.

Other things that seem to contribute to success with agents are:

- Static type systems (not tacked-on like Typescript)

- A test suite where the tests cover large swaths of code (i.e. not just unit testing individual functions; you want e2e-style tests, but not the flaky browser kind)

With all the above boxes ticked, I can get away with only doing "sampled" reviews. I.e. I don't review every single change, but I do review some of them. And if I find anything weird that I had missed from a previous change, I to tell it to fix it and give the fix a full review. For architectural changes, I plan the change myself, start working on it, then tell the agent to finish.

▲

BatteryMountain 2 hours ago | parent | next [-]

C# works great for agents but it works due to established patterns, strict compiler & strong typing, compiler flag for "Treat Warnings as Errors", .editorconfig with many rules and enforcement of them. You have to tell it to use async where possible, to do proper error handling and logging, xml comments above complex methods and so on. It works really well once you got it figured out. It also helps to give it separate but focussed tasks, so I have a todo.txt file that it can read to keep track of tasks. Basically you have to be strict with it. I cannot imagine how people trust outputs for python/javascript as there are no strong typing or compilers involved, maybe some linting rules that can save you. Maybe Typescript with strict mode can work but then you have to be a purest about it and watch it like a hawk, which will drain you fast. C# + claude code works really well.

▲

nl 3 hours ago | parent | prev [-]

Typescript is a great type system for agents to use. It's expressive and the compiler is much faster than rust, so turn around is much quicker.

I'm slowly accepting that Python's optional typing is mistake with AI agents, especially with human coders too. It's too easy for a type to be wrong and if someone doesn't have typechecking turned on that mistake propagates.

	▲	resonious 8 minutes ago \| parent \| next [-]
		Whenever I have an agent use Typescript, they always cast things to `any` and circumvent the types wherever convenient. And sometimes they don't even compile it - they just run it through Bun or similar. I know I can configure tools and claude.md to fix this stuff but it's a drag when I could just use a language that doesn't have these problems to begin with.
	▲	davidfstr 3 hours ago \| parent \| prev \| next [-]
		You should not be using Python types without a type checker in use to enforce them. With a type checker on, types are fantastic for catching missed cases early.
	▲	K0IN 3 hours ago \| parent \| prev \| next [-]
		Same for typescript, by default you still got `any`, best case (for humans and LLM) is a strict linter that will give you feedback on what is wrong. But then (and I saw this a couple times with non-experienced devs), you or the AI has to know it. Write a strict linter config, use it, and as someone with not that much coding knowledge, you may be unfamiliar and thus not asking.
	▲	embedding-shape 25 minutes ago \| parent \| prev [-]
		> I'm slowly accepting that Python's optional typing is mistake with AI agents, especially with human coders too. It's too easy for a type to be wrong and if someone doesn't have typechecking turned on that mistake propagates. How would you end up using types but not have any type checking? What's the point of the types?