正在加载视频...
视频加载失败
I'm open-sourcing /first-reader, a free agent skill that reads your draft the way an actual person would. first-reader simulates two people going through your draft one passage at a time, and reports how they felt - what's interesting, what doesn't make sense, things they skipped and what's something they... show more
11,483 次观看 • 26 天前 •via X (Twitter)
22 条评论

Thanks for shipping it!

You're welcome.

Love this. Super neat!

thank you!

Great job. Congrats!

Thanks!

my friends just say looks good

say thanks to your friends :)

like this coz ai feedback is way more useful when it focuses on how a real reader actually reacts, not just grammar and structure.

Neat idea!

thanks. glad you liked it.

passage-by-passage catches the moment a reader's mental model diverges.

Love the dual-reader simulation angle. Most review agents score clarity in one pass - forcing two perspectives passage-by-passage catches the makes-sense-to-me but confusing-to-them failure mode.

thanks

The “where they bailed” signal is way more useful than a generic writing score. Seeing exactly where attention drops could make editing much more actionable.

The passage-by-passage part is what I want. A whole-document critique lets the model smooth over the moment it got confused after seeing the ending. Showing exactly where each reader leaned in or bailed is much harder to fake with a polite summary.

The passage-by-passage part is the useful bit. I write the complaint down verbatim before anyone fixes it; otherwise the team optimizes for the clean summary and quietly loses the reader's actual snag. Curious what the simulated readers disagree on most?

I'm going to build this into my AI editor plugin for Obsidian ❤️

go for it

Writing to please a simulated attention metric risks training authors to write for AI quirks rather than real human readers. How do you prevent the feedback loop from optimizing for LLM biases?

Really interesting approach. Capturing how an agent moves through content, what it skips, and what actually drives its response is a powerful step toward understanding agent behavior. The next layer is making the underlying context and evidence just as traceable. 👏👏

The reader-emotion angle is the part I want. Most tools grade your text; this one reports how someone actually felt reading it.

