[HerbManic]

I like that it doesn't block the content but merely highlights it. That is a smart move.

The limit ultimately will be how well the algorithm can keep up with changes in LLM cadence over time. This is usually were project like this come undone, the concept it easy enough to build, it is the up to date data set where the real magic is.

But other than that, very cool to see and interested to see how it goes.

[bigger_fish]

Thanks! It's actually dynamic. The backend contains a sequence that checks daily for new model updates and human-published documents regarding AI writing tells, then adjusts scoring accordingly.

That scoring methodology is the only part of the extension that's not open source; it's hidden behind a Cloudflare worker to prevent reverse engineering.