The submitted site is a short workplace etiquette page aimed at a behavior many people now see daily: someone asks a coworker a question, the coworker drops that question into Claude or ChatGPT, then pastes the answer back as if they had answered it. The page says this is lazy, shifts the burden of comprehension to the recipient, and strips out the actual value the human was supposed to add, namely judgment, context, and a point of view. It recommends a simple rule instead. Use AI if you want, but read the output, validate it, and then answer as yourself.
That basic norm landed with a lot of people, especially those dealing with AI-pasted
Slack messages,
PR review replies, customer support responses, and long internal docs. The strongest practical framing was not "AI writing sounds bad" but "copy-pasting it tells everyone you may not have understood it yourself." Several people turned that into lightweight policy language for teams: own the output, disclose when something is unreviewed, do not flood coworkers with model-generated filler, and treat writing as part of the thinking process rather than a formatting step. A recurring theme was that AI has amplified an old failure mode. Before, people pasted
Stack Overflow answers, Google hits, or vague links they had not really digested. Now they paste far more text with far more confidence.
The biggest wrinkle was that many readers thought the site itself sounded AI-generated, especially the "angry" variant. That irony became part of the story rather than a side note. One camp took it as hypocrisy and proof that once your prose trips the "Claude voice" detector, the trust is already gone. Another camp pushed back that this has become a social panic of its own, where people overfit on em dashes, short punchy sentences, or certain stock phrases and start accusing humans of being bots. Either way, the credibility problem is now bigger than literal copy-paste. People are reacting to the texture of the writing itself.
The most useful disagreement was about context. Some people argued the page overstates "the other side has the same tools you do." In real work, your model may have access to repos, ticket history, customer details, internal docs, or a role-specific harness that the other person does not. In those cases, raw model output can contain genuinely useful context. Even then, the sharper view was that this only helps if the sender curates it. A human précis plus clearly labeled AI appendix was seen as acceptable. Blind paste was not. Several people also said overlong AI output is sometimes preferred to cryptic one-liners like "x broken," but only because the true underlying issue is poor communication discipline, not because model prose is inherently good.
The thread also widened into workplace norms around questions themselves. Some argued that pasting AI answers is just the 2026 version of "let me Google that for you" and a predictable response to lazy, low-context questions. Others rejected that framing and said human collaboration, mentoring, and documentation are part of the job. The practical middle ground was strong: ask better questions, state what you've tried, keep answers concise, and do not substitute model output for understanding. That is where the discussion settled. AI is fine as an aid. The sender still has to think, verify, and compress.
dontpastetheai.com