I spoke about jns' detection of gopher scraping by yandex,
pillaging the gopher to fuel their AI-deintellectual-
propertised proprietary generative bots. jns mentioned I
might lisp-hack a solution,
fortunately that knowledge is inside the OldComputerChallenge.
In this case I am drawing on writing from Interlisp (= the
Palo Alto Research legacy).
The interlisp commentary on common lisp was that it was a bland
portability layer. It lacked the personality and quirks which
were also interlisp's power. However there was the virtue; the
lack of character meant it was easy to downstream common lisp
projects into idiomatic interlisp, while it was very hard to
strip idiomatic interlisp fork extensions upstream to common
lisp.
This connects to discoveries about generative AI code-stealers.
They work best for bland tasks for bland, widespread targets.
Further, getting useful results has been discovered to be
culturally sensitive: The bot will respond to queries that
culturally match high quality source material with high quality
(scraped) answers.
An MIT discovered hack was querying- "who produced source material
for this goal" A: "jns". Q: "How would jns code this"-
which looks a lot like going to linkerror.com and lifting some code
in the first place.
This implies the stylish solution of forking bland projects rather
than becoming a user. We know that it's possible to characterfully
fork a bland codebase into a sm0l environment, but it is known to
be difficult to bring changes back upstream.
Commons, then, amount to a public good rather than a snare-trap to
create permanently dependent users, essentially a collection of
downstreamable performance and security ideas.
So don't become a customer of bland, mass-market commons: Fork them
in highly localised ways. This makes publically visible information
indigestable for modern generative AI, leads to sm0ller and self-
contained, specific codebases.
Furthermore more idiomatic and mutually aware programs work better,
like the idiomatic interlisp lispusers niche. There is also no need
for a portability layer to ingest decaying, bland non-local-source
codes.
Aside: The need to manually downstream security patches matches the
need to specialise pathogenic infections.
