laanwj on Nostr: maybe ? i'm personally skeptical of use of AI for technical documentation. It ...
maybe ? i'm personally skeptical of use of AI for technical documentation. It primarily needs to be precise, not verbose and flowery. Whereas the models often hallucinate things, from entire functions to subtleties like data types, field names, constraints on use. Anything a LLM would come up with would have to be reviewed with a dust comb. That's so much work that the people reviewing might as well write the documentation in the first place, and have a bit more confidence in it.
Published at
2024-09-08 07:02:24Event JSON
{
"id": "9867b150c40e3557310eb0b6d499c107b9032ebe117e18171987ef26547d691e",
"pubkey": "0aa39e5aef99a000a7bdb0b499158c92bc4aa20fb65931a52d055b5eb6dff738",
"created_at": 1725778944,
"kind": 1,
"tags": [
[
"e",
"c4c8f92d6a40bf382747d3561545d9578e7b4c2f5c8ea309ea235488597b1818",
"wss://nostr.wine/",
"root"
],
[
"e",
"4381cace65226050e97a08d5368059f3222c41c2bcbad3b1cea260b1729de011",
"",
"reply"
],
[
"p",
"12bd0599f099bf71e4f04589caf6b377f7700834c3d6f84e94825a8c68775357",
"",
"mention"
],
[
"client",
"noStrudel",
"31990:266815e0c9210dfa324c6cba3573b14bee49da4209a9456f9484e5106cd408a5:1686066542546"
]
],
"content": "maybe ? i'm personally skeptical of use of AI for technical documentation. It primarily needs to be precise, not verbose and flowery. Whereas the models often hallucinate things, from entire functions to subtleties like data types, field names, constraints on use. Anything a LLM would come up with would have to be reviewed with a dust comb. That's so much work that the people reviewing might as well write the documentation in the first place, and have a bit more confidence in it.",
"sig": "7ba7ed35692dbaa2063da25c9e9dd52e0641d59344753e70e104b618bc56024bf0dc7cbc2f915fe63802cd7ebdaf04ae3ab298123ab76f4d3a01cf992caf9aaf"
}
