← Glossary

Slop Grenade

The workplace anti-pattern of copying a coworker's question into an LLM and pasting the raw, unread output back at them instead of an actual answer -- a thoughtless lob of unverified AI text that the recipient now has to clean up.

Context

The term comes from noslopgrenade.com, which made the rounds in May 2026 and got covered in Dan’s rant on Episode 27 of the ADI Pod. The scenario is specific: a coworker asks you why you chose a particular architecture four months ago, and instead of answering, you paste their question into a codebase with Claude running and lob the raw model output straight back at them. You haven’t read it, you haven’t summarized it, you haven’t added your judgment — you’ve thrown a slop grenade.

The site frames itself as the spiritual successor to nohello.com, the long-running etiquette PSA against sending a bare “Hi” in a DM and then making the other person wait for the actual question. Both name a small, common rudeness that the medium makes easy: nohello externalizes the cost of starting a conversation; the slop grenade externalizes the cost of reading and verifying AI output.

Why It Matters

The slop grenade is an etiquette problem that doubles as a trust problem. The whole point of asking a human is that you want their judgment — their one-line take, their reasoning, the “plus my professional opinion” subtext that an LLM dump strips out. Pasting unread output says, in effect, “I valued your question so little that I wouldn’t even read the answer before forwarding it.” Worse, the recipient now inherits the verification burden the sender skipped: they have to figure out whether the wall of generated text is correct, which is often more work than answering from scratch would have been.

It has a sibling failure mode the episode flagged: the “let me Claude that for you” move, where the asker could have pointed the model at the repo themselves — the AI-era descendant of “let me Google that for you.” Both are about who does the work of turning a tool’s output into an answer.

The fix is not “never use AI to help respond.” It is to apply the same courtesy that already governed the medium: lead with your one-line take and a couple lines of reasoning, then attach the full output for anyone who wants the complete kaboom — and flag that it’s AI-generated. On the show, Shimin’s proposed UX patch was a chat-client toggle for “show my AI research,” analogous to a model’s show-thinking view, so the take and the supporting generation are visibly separated rather than mashed together.

Related Episodes