I think we are seriously overestimating the power of LLMs especially when it comes to building apps. Even the best models introduce subtle errors that can go unnoticed if you aren't paying deep attention. Just asking the model to generate tests should be scanned extensively with a fine-tooth comb.
if the goal is to build a throw away prototype sure full steam ahead. But for something that would be maintained and has more reach outside just self usage, this is a disaster in the making if not extra vigilant.
Xtr3m3hodl
npub174…mfau3
2025-06-06 02:47:39
in reply to nevent1q…cggj
Author Public Key
npub1749epjq95kgdher4x4rs3vpnjrthh25kryzntd9c55s6gzd4vzrqzmfau3Published at
2025-06-06 02:47:39Event JSON
{
"id": "7b95a291f5232397ec0f739a7180aa20517d980fff32c131c33296701cc276a4",
"pubkey": "f54b90c805a590dbe475354708b03390d77baa96190535b4b8a521a409b56086",
"created_at": 1749178059,
"kind": 1,
"tags": [
[
"e",
"6e27e49ac850afbc04a177b86fa21b9aa452f364f4452df0960ec9dc865bed5a",
"",
"root"
],
[
"p",
"932614571afcbad4d17a191ee281e39eebbb41b93fac8fd87829622aeb112f4d"
],
[
"p",
"460c25e682fda7832b52d1f22d3d22b3176d972f60dcdc3212ed8c92ef85065c"
]
],
"content": "I think we are seriously overestimating the power of LLMs especially when it comes to building apps. Even the best models introduce subtle errors that can go unnoticed if you aren't paying deep attention. Just asking the model to generate tests should be scanned extensively with a fine-tooth comb.\n\nif the goal is to build a throw away prototype sure full steam ahead. But for something that would be maintained and has more reach outside just self usage, this is a disaster in the making if not extra vigilant. \n",
"sig": "80f5895640c67f6bb05df16a2a6cc0625193f44d7d2e3c4a084002109c982dbd5542ff5385402592d89c67ef97455299acb712ed5ec0c5633b634ee7ed93d28c"
}