Why I built a one-keystroke rewriter

For a long time my evenings had a leak in them. The leak was usually a two-line reply. I would know exactly what I wanted to say, often within seconds of reading the message, and then I would write it five different ways, sit there comparing the versions, and send the fifth one an hour later. The content never changed between drafts. Version one and version five said the same thing. What changed was the temperature.

That gap bothered me more than it probably should have. The thinking was done. The decision was made. And still the message sat there, because the sentence that sounded fine in my head looked wrong on the screen. Too blunt in one version, too careful in the next. I would fix the bluntness and create a new problem two words later.

Unvent exists because of that gap. This is the honest version of why.

Thinking was never the slow part

The standard story about writing help goes like this: people struggle to produce words, so give them a machine that produces words. Blank page, writer's block, the cursor blinking accusingly. I don't buy that story, at least not for the writing most of us do all day. Nobody has writer's block in a chat box. The words show up instantly. What doesn't show up is permission to send them.

The real bottleneck is translation. You are holding a thought that is already complete and trying to convert it into something a specific reader can take in without a specific misreading. "No, Monday is realistic" has to become a sentence that keeps the no and still lands as friendly, sent to someone who cannot hear your voice or see your face. That conversion is genuine work. For short messages it is the only work. Drafting was never the hard part. Committing to a tone is.

You can watch this in yourself. Notice which messages you postpone. It is never the long ones. It is the two-liner to someone whose reaction you care about, where every word choice is a small bet on how they'll read it. The message is short precisely because the stakes live in the tone, and tone is the one thing you cannot proofread, because you already know what you meant.

So most writing tools optimize the wrong step. They help you produce a draft, which you already had, and then leave you alone at the exact moment you were actually stuck: staring at a finished draft, unable to press send.

Everything wanted a conversation

I tried the obvious fixes. The routine went like this: select the draft, copy it, open a new tab, paste it into an assistant, explain what I wanted, wait, read the result, wince at parts of it, copy the good sentence, switch back, paste, then re-read the whole thread because by now I had forgotten the exact question I was answering.

The tab is the first problem. Leaving the text box costs far more than the seconds it takes, because you come back a slightly different person. Whatever momentum the message had is gone, and you rebuild the context from scratch.

The conversation is the second problem. Chat assistants want to chat. They ask what tone you're going for, offer options with headers, tack on a cheerful closing line you never asked for, and invite you to iterate. Iterate! I was trying to escape a rewriting loop, and the fix on offer was a rewriting loop with an enthusiastic partner. You end up drafting prompts about your draft, which is the same disease one level up.

Grammar tools sit closer to the text, and I respect them for that. But they answer a question I wasn't asking. They will defend your commas to the death and go silent on the one thing that kept the message unsent: how does this sound to the person reading it?

Vent first, send the good version

The workflow I wanted was almost embarrassingly simple. Let me type the honest version first, straight into the real text box, with the irritation still in it. That draft is fast because it's true. Then let me pick the register: polished for the client, direct for the teammate who prefers it straight, warm for the person who deserves more warmth than I can type at 11pm. Then send, and go do something else with my evening.

Vent first. Send the good version.

That sentence is the whole idea, and it's where the name came from. You vent, and then you un-vent the message before it leaves. The venting is not a bug in the process. It is the fastest route to a complete draft, because the version with the feelings still in it is the version you can type without stopping.

For this to work, the tool has to live where the typing happens. Not an app you visit. Tone problems do not respect product boundaries: they show up in email, in tickets, in DMs, in the comment field of a doc at the worst possible moment. So it had to work in place, in whatever text field currently holds your cursor, and it had to leave you there when it was done.

And it had to be one keystroke away, which sounds like a detail and is actually the entire product. The window between "draft finished" and "send pressed" is a few seconds wide. A tool that fits inside that window gets used every day. A tool that misses it, even by one tab switch, gets used twice and forgotten. Speed is not a feature you add to a thing like this. Speed is the thing.

The vent draft also forced a decision I'm glad I made early. That draft is the rawest sentence you'll write all day, and it deserves paranoia. So drafts are never stored and never used for training, and the logging is content-free by design. I wrote those rules into the codebase as tests that fail if I ever drift. A promise in a README is a mood; one your test suite enforces is a policy.

An honest test

Unvent is my product, built the way I wanted it built. There is no roadmap committee to convince. When feedback lands, it reaches someone who can change the thing the same week. I like that more than I expected to.

I'm launching it free, on purpose. Not as a growth tactic, as an honesty tactic. Charging from day one would tell me whether my landing page persuades strangers, which is a fact about my marketing. A free launch tells me whether people come back on a Tuesday without being reminded, which is a fact about the product. I want the second fact first. If nobody wants this when it costs nothing, no pricing page is going to save it.

I should admit the part that keeps me humble. I built this for a version of me that I can't be sure is common. Maybe most people fire off two-line replies without a second thought, and my evenings were leaking through a hole most people don't have. I don't think so. I have watched too many capable people hover over a send button. But the free launch is how I stop guessing.

The replies that used to eat my evenings now take about as long as they deserve, which is under a minute. What I find out next is whether that was a personal repair or a public one.