Building a focused AI rewriting helper without killing the writer’s voice
I don’t need more ideas, I need less ritual.
A few months ago I realized something uncomfortable:
I don’t have a “no ideas” problem.
I have a workflow disease.
Most days looked like this:
- draft rough copy for a landing page or email
- feel that one sentence is clumsy
- copy it into an AI tool
- “make this clearer / shorter / less cringe”
- paste it back, fix formatting
- repeat this 20–40 times a day
On paper, that’s “being productive”.
In reality, my focus was getting shredded into tiny pieces.
At some point I stopped and asked myself:
“Why does every tiny rewrite require a full ritual with new tabs, new UIs, new distractions?”
That question is basically the origin story of what I’m building now: Rephrazo.
Where the pain actually was (it wasn’t “writing”)
If you zoom in on a single loop — copy → switch tab → paste → wait → paste back — it doesn’t look like a big deal.
But every time you do it, you:
- see a new interface
- see new notifications
- have to “reload” the context in your head
- come back to the doc half a step colder than before
One loop is fine.
Ten loops are annoying.
Fifty loops and the whole writing session feels like “managing tools around your text” instead of actually working on the text.
When I talked to other people (copywriters, founders, creators), they kept saying some version of:
“I know what I want to say. I just waste too much mental energy polishing it.”
That’s a very different problem from “please write this for me from scratch”.
Why yet another “AI writing assistant” didn’t solve it
I tried a bunch of things, including my own early prototypes:
- separate editors with beautiful UIs
- browser extensions with 10+ modes (formal, casual, punchy, etc.)
- side panels that wanted to become the new place where I write
All of them had the same flaw:
they wanted me to move into their world, instead of quietly fitting into mine.
At some point I wrote down a brutally simple requirement:
I don’t want “a new place to write”.
I want one very small action that just rewrites a sentence where it already lives.
No new window.
No prompt box.
No configuration panel.
Just: “this line feels off → fix it → keep going.”
That turned into Rephrazo (which is basically one gesture)
So I stopped thinking in terms of “product” and started thinking in terms of gesture.
What gesture do I actually want as a writer?
- highlight a sentence
- press one hotkey
- see a rewritten version appear right next to my cursor
- decide: insert it or ignore it
- continue typing
That’s it. That gesture is the product.
Everything else is infrastructure and error handling.
This is what Rephrazo does right now:
- you highlight any text
- hit a shortcut
- a tiny popup appears with an AI rewrite
- one click = text is replaced in place
No tab switch, no chat thread, no “what should I ask the AI this time?”.
The most surprising feedback: people don’t want “better”, they want “still me”
I assumed the most common request would be:
“Make this more persuasive / more punchy / more high-converting.”
That definitely shows up, but the strongest pattern is different.
What people actually care about:
- “Just don’t kill my voice.”
- “I’d rather keep my messy original than sound like generic LinkedIn AI.”
- “I want a clearer version of my own tone, not a different persona.”
So the internal “spec” for the rewrite changed from:
- “make it better” → to → “make it clearer, same person”
That tiny shift changes how people emotionally relate to the tool.
It’s not a replacement; it’s a helper that’s only allowed to sand the surface, not rebuild the object.
Options looked powerful. A strong default actually worked.
Of course, my first version of Rephrazo was over-engineered:
- you could choose tone (formal / casual / punchy)
- you could choose length (shorter / longer)
- you could see and tweak the underlying prompt
It felt “pro” to me.
Most users… ignored it.
What they did use:
- one main action
- one solid default behaviour
- almost no decisions
The more options I removed, the more often they pressed the hotkey.
It’s a bit painful to admit, but it’s true:
“flexibility” is often for the builder’s ego.
Users want one trustworthy move that works 80% of the time.
Where Rephrazo actually fits (and where it doesn’t)
Based on real usage and conversations, it seems to be most useful for:
- landing page sections
- outreach / sales emails
- X / LinkedIn posts and threads
- non-native English writers who “can write”, but want to sound a bit more natural
It’s not great for:
- generating first drafts
- big structural edits on long articles
- cases where the person has no idea what they’re trying to say
So I stopped pitching it to myself as “AI for writing” and started framing it more honestly:
“It’s a small, cheap way to rewrite sentences
once you already know what you want to say.”
Nothing more, nothing less.
Things I’m still figuring out (this is where Indie Hackers can probably help)
Right now I have more questions than answers:
- How do you know when a tool like this has crossed the line from “cool toy” to real habit?
- For something this narrow, what matters more early on:
– number of sessions,
– rewrites per session,
– retention,
– or something else entirely? - Where would you draw the line between “invisible UX” and “I want more control”?
Should there be modes again, or is one really good default the right hill to die on?
If you’ve built small, focused tools or AI helpers before, I’d love to hear what actually told you:
“Okay, this is no longer a gimmick — people are folding it into their real work.”
