Agreed about LLMs reinforcing errors. They aren't grounded the same way software like Mathematica are and can lead students astray.
For now I think they could be useful for people to explore ideas that lend well to verifiable feedback. For example, writing software that has a compiler or the user can manually inspect the output to see if it aligns with their vision. It opens the doors to try out things that we may otherwise forego because the expected payoff is too low.
MachuPikacchu / Machu Pikacchu
npub1r6…4gmmd
2025-11-04 17:36:01 UTC
in reply to nevent1q…y0cz
Author Public Key
npub1r6ggl0qazvwp02rlxgrf75lkfazuwhu35tmdg0u25eqsjax6243qh4gmmdPublished at
2025-11-04 17:36:01 UTCEvent JSON
{
"id": "91616f1da622451a1cfcc584fdf1a8dd0b8b9600d2e5fb22d60f1b78fff075e2",
"pubkey": "1e908fbc1d131c17a87f32069f53f64f45c75f91a2f6d43f8aa6410974da5562",
"created_at": 1762277761,
"kind": 1,
"tags": [
[
"e",
"122c97e5abc9b653962e54da2441680361f38046c5052e0fdd2e3e9dc335625e",
"wss://relay.nostr.band/",
"root"
],
[
"e",
"4625218bef9a6a8dfb1eda4bd515832a57acd47639d55f6459e4eb804e1e00e5",
"",
"reply"
],
[
"p",
"675b84fe75e216ab947c7438ee519ca7775376ddf05dadfba6278bd012e1d728"
]
],
"content": "Agreed about LLMs reinforcing errors. They aren't grounded the same way software like Mathematica are and can lead students astray.\n\nFor now I think they could be useful for people to explore ideas that lend well to verifiable feedback. For example, writing software that has a compiler or the user can manually inspect the output to see if it aligns with their vision. It opens the doors to try out things that we may otherwise forego because the expected payoff is too low.",
"sig": "4fc09cc14bdae6ebd6d3467e66e9732b82cc9f5de40489fff4c5e58d36c76d55170827c680f14f722c9e86e3162a9f3a1bb4fed529b753c0337c404479caa546"
}