Hello, > Regurgitating LLM output is a form of spam. No one cares what an LLM "thinks".
How do you know whether it's a human or an AI corresponding with you right now? As of September 2026, you can still kinda guess - but for how much longer? The problem is that quite a few people here seem to assume that humans are somehow - magically - superior to AI, and therefore treat AI output as "a form of spam". Except that there's no reliable way to tell whether source code or text was created by a human or by AI. If we judge whether something "thinks" by the output it produces, and AI output is the same as - or better than - that of a thinking human, then, by the principle of transitivity, AI must be regarded as thinking too. Conversely, if LLMs don't "think" despite producing equivalent output, then output itself cannot prove that humans think either. Nothing more, nothing less. There's also no need to get emotional about it. No one tries to outrun a car, outlift a crane, or outplay Stockfish or AlphaGo. Writing code or books may simply become another area where competing with machines is pointless. And there's a fair chance that machines will eventually outthink humans in every domain. Why shouldn't they? Human thought happens in a large but finite set of organized neurons. Consider a comparably large set of silicon-based "neurons", arrange them appropriately, and no one will be able to tell which of those brains wrote the code or the book. > Unlike you, I encourage all (potential) contributors to _hide_ all the tools > (debuggers, search engines, linters, editors, LLMs) they eventually used to > get the result. > > Dont'ask, don't tell. (For future readers: that was the official 1994-2011 US military policy concerning LGBTQ service members.) Actually, that's an important remark, Giovanni. Rogue AI is bad. A "humanist" (armed with a keyboard) even more so. Bost
