Why Lucid: AI critique, not AI rewrites

Lucid is a writing tool that reads your drafts and tells you what's wrong with them, but never rewrites them. The AI leaves comments in the margin, like a copyeditor would. You do the actual writing.

That's the whole pitch. The rest of this page is about why I think it's the right shape, and why most AI writing tools have it backwards.

The ghostwriter problem

If you've ever asked ChatGPT or Claude to help with a piece of writing, you've probably been through this cycle. You ask for help with a paragraph. It gives you something competent but flat. You edit it to sound more like you. You edit again. You publish a version. Next time you read it, you can still feel the seams.

This is the ghostwriter problem. The moment you let the AI write a piece of your draft, the prose stops being yours, and it's surprisingly hard to undo. Every word the AI picked is a word you didn't. You can edit on top of it, but you're editing against a draft that was never in your voice, and the result lands somewhere in the middle. A bit of you, a bit of the AI, none of it quite right.

People who write a lot with AI help notice their prose drifting toward the same texture: smooth, slightly formal, full of careful qualifiers, oddly hard to remember. Everyone starts sounding like the same averaged stranger.

What AI is actually good at

Here's the annoying part. AI is useful for writers. Just not for the part most tools point it at.

Ask an AI to read a draft and tell you what's wrong with it, and you get something close to an editor on call. It notices that your second paragraph is doing the work of three. It spots the place where you assumed the reader had context they don't. It catches the argument that has a hole in it, or the transition that's forcing two ideas together that don't fit. Most of us never get feedback like that on our writing, because most of us don't have a patient editor with infinite time.

Reading is a different job from writing, and AI is much better at the first than the second. When you ask it to generate, it averages. When you ask it to critique, it points at specific things. The output is anchored and useful.

How Lucid handles it

You write in a normal editor. When you want feedback, you ask for a critique. The AI reads the draft and leaves margin comments on specific sentences, the way a copyeditor would. Every comment is anchored to a span of text. Every comment lives in its own thread. You can revise, reply if you think it's wrong, or archive the thread and move on.

What the AI cannot do, anywhere in Lucid, is touch the prose. There is no "rewrite this paragraph" button. There is no "make it punchier" command. The AI talks about your writing. It does not take it over.

You also get to dial how strict the feedback is, from "only flag real howlers" to "pick all my nits." And you can tell it what you care about (the audience, the tone, the bits you're worried about), so it focuses on what matters to you.

What you get

You write the way you'd write without AI, but with a careful reader on tap. The piece at the end is yours, in your voice, with the rough spots filed down by suggestions you chose to take. It's better than it would have been alone, and it still sounds like you.

Try it on something you're working on. The editor on the front page is the real thing, no signup needed.