AI Pulse · Under the Radar

If you’ve used Claude or any agent-based tool to draft customer emails, proposals, or reports, you’ve probably noticed a certain sameness creep in. Phrases like “it’s worth noting” or “in today’s fast-paced world” show up so often they’ve earned a name in AI circles: slop. It’s not wrong exactly, it’s just detectable. Readers can tell when a human didn’t touch the final draft, and that costs you trust in ways that are hard to measure but easy to feel.

A small tool called sloptrim is trying to fix this at the source. It’s a plugin for Claude Code that runs in the background, checking generated text against 71 documented patterns of AI writing tells, then flagging or auto-fixing the spans it catches. The part worth paying attention to is that it runs fully local. No network calls, no sending your drafts to another server to get scored. It just sits on your machine and cleans up as you go.

Right now it’s tiny, sitting at 39 stars on its repository, so this is early and unproven at scale. But the problem it’s chasing isn’t niche. Anyone running content, sales outreach, or client communication through AI agents has hit this same wall: the output is fast, but it doesn’t sound like a person wrote it, and that shows up in reply rates and client trust. Tools like this are a signal that the next wave of AI tooling isn’t just about generating more text, it’s about making that text pass as genuinely yours.

For business owners layering multiple AI tools into daily operations, this kind of quality-control step is exactly the sort of thing worth building into your workflow rather than bolting on later. It’s the kind of thing we build into an AI command centre, so the polish happens automatically instead of becoming one more manual review task on your plate.