Glonce

Why hobby programming communities are hostile to LLMs

miscThe Glonce editorsHacker News🔥58Glonce rating58Impact28Novelty42Relevance88Surprise35Players12Credibility72

A post dated August 4, 2026, titled "Born Against, or why hobby programming communities are against LLM usage," on the blog blog.fogus.me by an author identified as Fogus, who signs off as ":F," argues that niche hobby-programming communities are increasingly hostile to LLM use. The piece was prompted by a GitHub thread about chess-engine development, though the author notes the thread itself does not explain much; it just prompted the reflection. The author reports seeing the same hostility in several other narrow, skill-driven hobby-programming scenes: OSDev, LangDev, TxtDev, EmuDev, RLDev, the demoscene, and code golf.

The essay's central claim, framed as a general pattern rather than a universal rule (the author explicitly flags an implicit "not all" throughout), is that the knowledge built in these communities is hard fought, which makes using an LLM "a form of missing the point entirely." In these scenes, the author writes, "the process of mastering a difficult field itself is the product, and something that runs is generally a nice-to-have." Respect is earned slowly, through years of activity in a community's own forums, by sharing elegant code, showing genuine curiosity, and demonstrating deep domain knowledge. Whether the code works at all is beside the point; what these communities care about, per the author, is whether the person knows why and how it works.

Even where there was earnest, early engagement with LLMs inside some of these communities, the author writes that "the well was quickly poisoned" by two things combined: a lack of deep understanding among the people using LLMs, and a vitriolic subset of the community that treats using an LLM as a form of cheating. The author connects this to the communities themselves, which have, in general, "been characterized by feverish gatekeeping and painstakingly slow progress," creating what the essay calls a temptation to grab "easy cachet by bursting onto the scene like the Kool-Aid man." The line right after deflates it: "OH YEAH.... OH NO!"

The author's own view of LLMs is more measured: an LLM "functions best as a force multiplier, not a surrogate," and "in the hands of an expert who already understands a domain deeply, it could act like a lever." But in these specific niche communities, the author argues, the entire exercise is the learning itself, so using an LLM to generate the finished piece "doesn't make us craftsmen; it just robs us of the craft."

A footnote adds a separate caveat: "expertise offers no natural immunity against being fooled by LLMs." The post is presented as the latest entry in an ongoing series of the author's reflections on LLMs, following two earlier pieces it links back to and names as "LLMe" and "Mind the van Emden Gap."

Key facts

  • Dated August 4, 2026 on the blog blog.fogus.me, titled "Born Against, or why hobby programming communities are against LLM usage," identified as Fogus, and signed off as ":F."
  • The post was prompted by a GitHub thread about chess-engine development, then extended to OSDev, LangDev, TxtDev, EmuDev, RLDev, the demoscene, and code golf.
  • Its central claim: in these communities mastering a hard field is the actual product, so an LLM-generated result that merely runs is "a nice-to-have" and treated as "a form of missing the point entirely."
  • Even earnest early attempts to use LLMs in these communities soured, per the essay, from a mix of shallow understanding among LLM practitioners and a vitriolic faction that calls LLM use cheating.
  • The author's own verdict: an LLM works best as "a force multiplier, not a surrogate," useful in the hands of an expert as "a lever," but generating the finished piece with one "just robs us of the craft."

Why it matters

The essay names a specific, checkable fault line rather than a vague claim that developers dislike AI: in communities built around chess engines, OSDev, LangDev, TxtDev, EmuDev, RLDev, the demoscene, and code golf, a working LLM-generated result does not earn the status a hand-won one does. That matters because it cuts against the assumption that AI coding tools are a universal upgrade every programmer welcomes; in pockets where the skill itself, not the output, is the actual currency, the opposite is happening.

Who it affects

Directly, hobbyist programmers in niche, skill-driven fields: chess-engine developers and the OSDev, LangDev, TxtDev, EmuDev, and RLDev communities, plus the demoscene and code golf, groups the author describes as long defined by "feverish gatekeeping and painstakingly slow progress." It also affects anyone hoping to use an LLM to gain quick standing in one of those scenes, since the essay says that shortcut reads as cheating rather than achievement, and more broadly anyone assuming technical hobbyists are an easy adoption base for LLM-assisted coding.

How to use it

The essay's own practical takeaway is about how, not whether, to use an LLM: as "a force multiplier" once you already understand a domain deeply, the way a lever multiplies force that already exists, not as a surrogate that hands you a finished result you could not have produced yourself. In the communities described, arriving with LLM-generated output you cannot explain will not read as an accomplishment, since what gets judged is whether you know why and how the code works, not whether it runs.

How solid is it

This is a first-person blog essay, not a study. No survey, poll, or external data source backs the "general consensus" it describes, which the author presents as a personal impression built from watching several communities over time. The piece names no specific example, thread, or individual for the "well was quickly poisoned" episode, does link to the GitHub chess-engine thread that triggered it, names no particular LLM, model, or company, and gives its own author's name, Fogus, beyond the signature ":F." It is also explicitly one entry in an ongoing personal series on LLMs, following two earlier posts it links back to. On Hacker News, the piece reached 170 points and 166 comments within about 10 hours, a strong discussion response, though that measures attention rather than independent verification of the argument.

Risks and caveats

The author flags an implicit "not all" running through the whole argument, so this is a described pattern, not a claim that every member of every listed community rejects LLMs. The essay also undercuts its own gatekeeping explanation with self-aware humor, deflating its own Kool-Aid man comparison a line later, which suggests the author does not treat that explanation as the full story. A footnote adds a further caveat that cuts against reading the piece as simply anti-LLM: "expertise offers no natural immunity against being fooled by LLMs," meaning even the deep domain knowledge these communities value would not, on the author's own telling, protect someone from being misled by one.

“Using an LLM to generate the finished piece doesn't make us craftsmen; it just robs us of the craft.”

— the author, signed ":F," on blog.fogus.me