No Meat Proxy logo

No Meat Proxy

AI is a tool. You shouldn't be.

Don't paste AI generated text as your answer

✕ Don't
Jim
Jim
Hey, how does the feature flag system decide which users see the new checkout flow?
Kevin
Kevin
You can find the logic in feature_flags/rollout.py, in the evaluate_rollout() function at line 142. It builds a composite key from the user_id and flag_key, hashes it with murmur3, and checks the result modulo 100 against the rollout_percentage field on the FeatureFlagConfig model. Results get cached in Redis for 5 minutes to avoid re-evaluating on every request, and there's a fallback path in evaluate_rollout_legacy() that kicks in if the cache lookup times out, which uses the same hashing scheme but reads the percentage from the old feature_flags.yaml config instead...
✓ Do
Jim
Jim
Hey, how does the feature flag system decide which users see the new checkout flow?
Kevin
Kevin
It's basically a weighted coin flip tied to your user ID, so you always land on the same side. Right now it's set to 20%, so about 1 in 5 people get it.

When someone asks you a question, they're asking because you're supposed to know the answer. They trust you and your expertise. Pasting back whatever the AI said leaves them wondering whether you even read it, and whether it's actually correct.

It's also disrespectful. It's not far from replying with a “let me google that for you” link: they asked an honest question, and you couldn't be bothered to give them a real answer.

If professionalism and respect aren't reason enough, consider this: pasting AI answers like this advertises how replaceable you are. Why would anyone need you if all you do is forward an AI's output? At that point, they could just cut out the middleman and ask the AI directly.

If you're an AI user, then use AI to derive at an answer, but don't offload the burden to decipher and verify that answer onto whoever asked you.

Remember: they asked you, not ChatGPT.

Don't blame the AI for what you signed off on

✕ Don't
function chargeCustomer(order) {
+ try {
+ processPayment(order);
+ } catch (err) {
+ logger.warn(err);
+ }
Jim
Jim
Why are we swallowing this error instead of letting it propagate?
Kevin
Kevin
lol, idk, Claude wrote it.
}
✓ Do
function chargeCustomer(order) {
+ try {
+ processPayment(order);
+ } catch (err) {
+ logger.warn(err);
+ }
Jim
Jim
Why are we swallowing this error instead of letting it propagate?
Kevin
Kevin
Upstream retries on any thrown error, and this one's expected during backfills, so we log and drop it here to avoid a retry storm. Can add a comment explaining that.
}

Even if AI did most of the work, you're still responsible for it. You should be able to answer basic questions about your own output, code or otherwise. Handing that responsibility to an AI is never acceptable.

It's also disrespectful. Not bothering to do the work yourself, then pushing the burden of verifying it onto someone else, just shows how little care you have for the people you work with and the work you do.

If you use AI, you still have to understand and verify what it gives you. You don't get to wave away responsibility. Blaming the AI just tells people not to trust your work.

Offloading responsibility to AI highlights how replaceable you are. If all you do is prompt an AI and submit whatever comes out, why do they need you?

"You're just an AI hater"

If you use AI, more power to you. This page isn't about demonising AI usage. Heck, this very website is partly AI generated (I'm terrible with CSS). If all you got out of this page is "AI hater", then you're the problem.