"Workslop" describes the problem of colleagues dumping AI-generated text on you without effort. The core issue is asymmetry: it takes seconds to generate with an LLM but still costs you real time to read and process. The author lays out five practical responses depending on your situation. If you have authority, just tell people to stop. If not, you can treat them like a coding agent interface—redirect their outputs back through Claude to force some quality control. You can also fight fire with fire by pasting their messages into your own LLM to extract key points, or ask an AI to draft responses for you. The trade-off here is you become part of the problem, but it beats spending ten minutes on every ten seconds of their effort. The social solutions work better in some contexts. Pushing conversations to calls or in-person meetings kills workslop at the source since you can't send AI text over a voice channel. More importantly, forcing someone to spend their own time talking to you makes the effort symmetrical again, which filters out people who are just being lazy or predatory. For lower-stakes stuff—status updates, external pull requests, organizational reports—you can simply skim or ignore it. Match their lack of effort with your own. If something actually matters, they'll explain it in their own words. The author includes a nuance worth noting: not all AI-generated content counts as workslop. If someone genuinely spent time crafting, editing, and thinking through their message even if it's AI-assisted, that's different from just copy-pasting a raw model output. The real problem isn't AI text existing—it's the deliberate offloading of effort onto someone else. Workslop is just a specific case of bad communication made worse because the barrier to sending it is so low.