are all the tests hand written or are some agent-contributed? curious

What’s the difference if you review the code getting merged?

Reviewing the correctness of code is a lot harder than writing correct code, in my experience. Especially when the code given looks correct on an initial glance, and leads you into faulty assumptions you would not have made otherwise.

I'm not claiming AI-written and human-reviewed code is necessarily bad, just that the claim that reviewing code is equivalent to writing it yourself does not match my experience at all.

Plus if you look at the commit cadence there is a lot of commits like 5-10 minutes a part in places that add new functionality (which I realize doesn't mean they were "written" in that time)

I find people do argue a lot about "if it is reviewed it is the same" which might be easy when you start but I think the allure of just glancing going "it makes sense" and hammering on is super high and hard to resist.

We are still early into the use of these tools so perhaps best practices will need to be adjusted with these tools in mind. At the moment it seems to be a bit of a crap shoot to me.

eh with plenty of tests that I can easily read and are well documented I haven't actually ever found this to be a problem in practice

Code review basically never actually means thinking through all the code again as if you wrote it

The difference is we can't tell if you reviewed the code.

To be fair, we also couldn’t tell for sure if they hand-wrote the code.

If they hand wrote the code we know they at least looked at it once.

Since we can’t know whether they really hand-wrote the code, we also wouldn’t know whether they looked at it.

Err yeah that's the point.

You wrote: “If they hand wrote the code we know they at least looked at it once.”

But that’s not true, because even if they did indeed hand-write the code, we as third parties wouldn’t have any reliable proof of that, and therefore still couldn’t draw the conclusion that they looked at the code.

Them claiming to have hand-written the code isn’t any better than them claiming to have thoroughly reviewed the code. We can’t know in either case.

> Them claiming to have hand-written the code isn’t any better than them claiming to have thoroughly reviewed the code. We can’t know in either case.

It is better, because most people aren't out-right liars. If he said "I hand-wrote the code", sure it doesn't prove it, but I would believe him. When he says "I thoroughly reviewed the code"... yeah maybe. That's the sort of thing people do lie about (even to themselves).

you never can, thats the responsibility of the engineers to not lie. You can tell if there are sufficient tests and if they passed

i mean idk that's sorta like asking what's the difference of having tests if you review the code getting merged

Did you actually review 313,397 LOC written by claude? And you wrote the tests? That's honestly very impressive if yes.