Okay real research question (I'm curious about y'all's thoughts)--the lowest common denominator of the genai/vibe coding debate in education is something like "regardless of how the code gets written, good computing education ensures students are able to understand and explain whatever they generate". Now that's not the end of _my_ goals an an educator, but it's the part of it I share with the vibe-coding boosters.
So then the real question is: how easy is it to learn to read & understand code by _writing it_ vs _generating_ it? If there's a transition point (e.g., student write simple programs & generate complex ones), where is that and how do we know when to make the transition? And what's the difference for classroom students vs informal, self-taught learners (this is the area that I get really animated about--LLMs are having a profound regressive effect for informal learners imo).
For coding educators, it doesn't matter how good LLMs are at generating code--our novice students will always be worse, but that's normal and fine. What matters is how it facilitates or inhibits the skill acquisition that make reading & analyzing code possible. Maybe the industry doesn't care about individual speed anymore bc genai is faster (it shouldn't ever have, imo), but I'm trying to get people to _learn to code_, not _produce software_. And I'm not convinced genai has a role to play that outstrips the pedagogy that predates it, at least not yet (and that's beside all my other objections to genai, which are numerous).
octojon
npub19h…t9a4e
2026-02-20 12:41:48 UTC
Author Public Key
npub19h9sa6v6ywdqsgzkusc75uypqk9mekmypjk5lk0qyjnrcpyp635spt9a4eSeen on
wss://relay.momostr.pinkPublished at
2026-02-20 12:41:48 UTCEvent JSON
{
"id": "ab2597eccb665906e1de566ddc076da9c1d4637c8f6d838684cc114c1f18f086",
"pubkey": "2dcb0ee99a239a082056e431ea7081058bbcdb640cad4fd9e024a63c0481d469",
"created_at": 1771591308,
"kind": 1,
"tags": [
[
"proxy",
"https://merveilles.town/@octojon/116103008006186849",
"web"
],
[
"proxy",
"https://merveilles.town/users/octojon/statuses/116103008006186849",
"activitypub"
],
[
"L",
"pink.momostr"
],
[
"l",
"pink.momostr.activitypub:https://merveilles.town/users/octojon/statuses/116103008006186849",
"pink.momostr"
],
[
"-"
]
],
"content": "Okay real research question (I'm curious about y'all's thoughts)--the lowest common denominator of the genai/vibe coding debate in education is something like \"regardless of how the code gets written, good computing education ensures students are able to understand and explain whatever they generate\". Now that's not the end of _my_ goals an an educator, but it's the part of it I share with the vibe-coding boosters.\n\nSo then the real question is: how easy is it to learn to read \u0026 understand code by _writing it_ vs _generating_ it? If there's a transition point (e.g., student write simple programs \u0026 generate complex ones), where is that and how do we know when to make the transition? And what's the difference for classroom students vs informal, self-taught learners (this is the area that I get really animated about--LLMs are having a profound regressive effect for informal learners imo).\n\nFor coding educators, it doesn't matter how good LLMs are at generating code--our novice students will always be worse, but that's normal and fine. What matters is how it facilitates or inhibits the skill acquisition that make reading \u0026 analyzing code possible. Maybe the industry doesn't care about individual speed anymore bc genai is faster (it shouldn't ever have, imo), but I'm trying to get people to _learn to code_, not _produce software_. And I'm not convinced genai has a role to play that outstrips the pedagogy that predates it, at least not yet (and that's beside all my other objections to genai, which are numerous).",
"sig": "89968ec36031303b8467a7a04068f396f4b599b04f3d6f97c9e55763e91c344ed40af5af3832f9aad21307508b749d3805158d2d838e0a8e8888998650f1dc01"
}