Chester Wisniewski learned to program by typing BASIC listings out of magazines into a Commodore 64, then by writing games for the BBS he ran. He traces how that world settled into a working balance: shareware and freeware first, then copyleft licenses — GPL, MPL, CC-SA — which used copyright itself to force openness downstream. Forty years of argument got us to an equilibrium. He argues LLMs have thrown it out, and that this loss gets far less attention than AI’s environmental or security problems.

What he says the models broke:

  • Training runs consume everything online “without regard for copyright or license,” and derived works may not carry the original license’s obligations — with, he says, no appetite for enforcement. His summary line: “The social contract has been broken.”
  • The incentive for an author now runs backwards. Publish your code and AI makes it easier to find and abuse your mistakes; keep it closed and those same mistakes are harder to find.
  • Publishing also invites a flood of AI-generated pull requests from bots and inexperienced users — “mostly useless slop” that eats the maintainer’s spare time in triage.
  • Credit and intent dissolve in the aggregate. His individual ethical guidelines, like the license terms, are lost once absorbed.
  • The consequence is the thesis: “AI has turned sharing knowledge from a gift to the world, into a liability for the author.” The previous 35 years of openness, he notes, built nearly everything we depend on online — the modern internet and cloud are derivative works standing on earlier generations’ shoulders.

The essay is a diagnosis with no mechanism attached. It points at “the colossus” — the companies doing the absorbing — and ends on a call for a “new digital Renaissance” from authors, coders and artists, warning of a “digital dark age” that reads like “some bizarrely twisted version of Fahrenheit 451.” What it does not offer is a licensing scheme, a legal theory, or a technical countermeasure. That gap is exactly what the thread argues about.

The 232-comment thread on Hacker News

What the thread adds

  • mmaunder — the working-maintainer rebuttal. He runs a 40-person team that makes its money shipping GPL software (WordFence), and pushes back point by point: “You have the same incentive you had to share your work as before, and that is to get attention, and customers, assuming that’s your game.” On the slop: “the tsunami of updates we need to do weekly is far more easy to handle with LLMs.” He concedes the unreversible part — “AI trained on your code and is selling that capability kind of sucks” — then lands on the trade: “we get AI! Which I frikkin love.”
  • grumbel — the strongest dissent, and not a small one: “AI is the best thing that ever happened to the Free Software world, it is basically turning any software into Free Software.” Throw file formats, protocols or plain binaries at a model and it can reverse-engineer them; users can finally modify software themselves, which was always the point of free software and rarely happened in practice. Their claim is that the problem the social contract existed to solve is gone: “Creation is now easy and commodified.”
  • archagon — a countermeasure rather than an argument. They’ve started attaching license addendums stating their projects are “not for LLM use or training,” explicitly to document intent should enforcement arrive later: “If and when the tides turn and these companies start getting sued, I want it documented that they were explicitly violating my license.”
  • chrka — identifies the motive the essay underweights: acknowledgement. A model can absorb an open-source project, rewrite it in another language, and ship it as its own product without applying the license or crediting the author. Their case is personal: “My former open-source project is now no longer open source.”
  • sigmar — corrects the essay’s premise that closing your code protects you. Frontier models decompile well enough that private source isn’t a shield — evidence offered is the run of clean-room game ports appearing lately. JoshTriplett’s one-line reply to the “clean room” framing: “Anything done with an LLM is not a ‘clean room’.”
  • mark_l_watson — a Creative Commons author who hadn’t read AI training as a license violation, and now thinks it is. Their practical response is to relicense: move their books from non-commercial share-alike to the CC variant that permits commercial reuse, plus a demand that scraping respect robots.txt and ai.txt.

Where the thread pushes back

  • Several commenters argue both of the essay’s stated grievances predate LLMs. luma: open source always let others review your code for their own purposes, and always produced low-quality pull requests — “AI changes absolutely nothing about this for me.” ricardobeat adds that the “no appetite for enforcement” line sits awkwardly against the essay’s own praise of open sharing. mmaunder makes the same move on vulnerabilities: “Open code has always been easier to find vulns in vs closed. AI didn’t change that.”
  • Forgeties79 answers that cluster with the scale argument, in what amounts to the thread’s best rebuttal: “Before LLM’s there was___” is the standard deflection, but the issue is that ever-smaller groups can now execute at scale. The analogy: “People could always die crossing a street. Still, cars changed the discussion about pedestrian safety pretty materially.”
  • japhyr supplies the reframe the essay itself won’t make. The sentence about AI throwing out 40 years of equilibrium should read “the companies building frontier AI models have thrown this out the window” — “We can’t hold a technology responsible for anything, but we can hold people responsible for their impact on everyone else.”
  • One sharp exchange is worth reading as a measure of temperature rather than of merit: tony_starkling calls the position “selfish and impractical” given the open code every career is built on, and bakugo answers — “There’s something deeply ironic about calling someone selfish for not wanting billion dollar corporations to suck up everything they’ve ever created and then sell it back to them.”

Reading note: HN handles are pseudonymous, the site publishes no per-comment scores, and the ordering here is HN’s own ranking — so this is a slice of the thread, not a consensus. Views above are attributed to the handles that wrote them.