(docs.astral.sh)

129 points NeutralForest | 2 comments | 03 Jul 25 11:51 UTC | HN request time: 0.4s | source

Show context

qoez ◴[03 Jul 25 13:10 UTC] No.44454665[source]▶

The comments in this thread all feel like auto generated engagement bot replies you'd see on twitter.

Ygg2 ◴[03 Jul 25 13:16 UTC] No.44454737[source]▶

Bots don't make that much spelling mistakes. Bots as in LLMs not people paid to engage.

My good uv experience. I tried installing tensor/cuda Python code recently. Plain pip just failed. uv pip actually returned WHY it failed.

It definitely felt like magic.

coldtea ◴[03 Jul 25 13:48 UTC] No.44455058[source]▶

>Bots don't make that much spelling mistakes.

They do if you instruct them to.

replies(1): >>44460086 #

Ygg2 ◴[03 Jul 25 23:52 UTC] No.44460086[source]▶

Why would you want them to make spelling errors? The benefits don't outweigh the costs.

replies(1): >>44464846 #

1. coldtea ◴[04 Jul 25 14:25 UTC] No.44464846[source]▶

If the intention is "increase believability a real human wrote this bot message" they do

replies(1): >>44465318 #

2. Ygg2 ◴[04 Jul 25 15:21 UTC] No.44465318[source]▶

Sure but adding error rate also makes it so you trust the speaker less, because they aren't educated enough or ESL.

The uv build back end is now stable