Why I Built Limato: From ChatGPT Round-Trips to Inline Rewrite

Mikhail Konkov · Founder, Limato

Russian-speaking developer living in Spain. Writes in English and Spanish daily. Got tired of the tab-switch.

For years, every important sentence I wrote — in English to US clients, in Spanish to my landlord and my doctor — went through the same loop: type a draft, copy it into ChatGPT, ask it to "make this sound native," wait, paste back, edit one more time. Multiply by thirty messages a day and the math stops working.

This is the story of how that loop became a Chrome extension — and why it's the right shape for the problem.

The years-long copy-paste loop

I'm not a beginner in English. I read in English, I can hold a conversation without translating in my head. But when I sit down to write — a Slack message to a US client, a LinkedIn post, a polite "no" to a recruiter — there's a small voice that says this sentence doesn't sound right.

It's almost always right, the voice. I'd write "I would like to kindly request your feedback" and feel the wrongness without being able to fix it. So I'd open a new tab, paste it into ChatGPT, and type the same prompt I'd typed five hundred times:

ChatGPT would return something like "Mind taking a look when you get a chance?" — exactly what I meant, in fewer words, without the ESL fingerprint. I'd copy it, switch tabs, and paste it back.

For a single message, that loop costs twenty seconds. For thirty messages a day — Slack, Gmail, LinkedIn DMs, PR descriptions, GitHub issues — it costs an hour. Every switch breaks whatever I was thinking about.

Why Grammarly didn't fix it

I tried Grammarly. I paid for Grammarly. Grammarly does what it advertises — commas, agreement, the occasional flagged passive — but it doesn't touch the underlying shape of an ESL sentence. It polishes "I am writing to kindly request that you revert back to me at your earliest convenience" into the same sentence with cleaner commas. The phrase is the problem. Grammarly defends the phrase.

What I needed was something that would rewrite the sentence — kill "kindly", kill "revert back", kill "at your earliest convenience" — and give me back the version a native speaker would have typed from scratch. Shorter, less polite, more direct. Something like:

That's a dramatic rewrite, not a grammar fix. Grammarly was the wrong tool. ChatGPT was the right tool wrapped in the wrong interface.

The shape of the fix

Eventually the answer was obvious. I didn't need a better model. The model I was already using was fine. I needed to remove the tabs.

I wanted to: highlight a sentence anywhere I was typing, press a shortcut, and have the rewritten version replace my selection. No new tab. No paste. No prompt. The rewrite happens where I'm already writing — Gmail, LinkedIn, Slack, Notion, Google Docs, a GitHub issue, this very text editor.

A Chrome extension was the obvious vehicle. Manifest V3, a small content script, a service worker that talks to an API, and a model on the other end that's been told to do exactly one thing: rewrite this text the way a native speaker would phrase it from scratch.

The hardest part wasn't the AI. The hardest part was the tone.

The "Native" tone is opinionated

Most rewrite tools are calibrated for safety. They keep your structure, polish your word choice, and ship you back a version that's recognizably your sentence with the rough edges sanded. That's useful for native speakers who want to tighten a draft. It's the wrong calibration for ESL writers, because the rough edges are the ESL fingerprint.

So Limato's Native tone is opinionated in a way I had to decide before writing the prompt:

• Kill ESL openers. "I am writing to", "I hope this email finds you well", "Kindly", "Please be informed that" — all gone, every time.
• Compress. Native rewrites are typically 30–50% shorter than the original. If the rewrite is the same length, the model didn't do its job.
• Prefer the verb. ESL English drowns in nouns and passives. Native English moves on verbs.
• Don't be polite. Be direct. Native business English is much less polite than ESL textbooks teach. "Could you" beats "I would be grateful if you could."

These are calibration calls, not grammar rules. They're the reason the output surprises people the first time — "wait, this is much shorter than what I wrote." That's the point.

Translate as a side effect

Once the rewrite pipeline existed, translation came almost for free. I'd write a draft in my native language, pick the same shortcut, and get an English version that's been through the Native-tone rewrite — not a literal translation, but the version a native speaker would have typed if they'd had the same thought.

This turned out to be the feature I use most. When I'm tired, I think in my native language. The translation isn't the point — the rewrite is. A literal translation inherits the source language's sentence rhythm. Limato strips that.

What it looks like in practice

Here's a before/after from this week — a Slack message I almost sent to a US client:

Same meaning, tighter. The ESL fingerprints disappear — "kindly", "wanted to check in regarding", "discuss about" — and what's left reads like a quick Slack message instead of a formal letter. My version reads polite-but-foreign. The rewrite reads like an American typed it.

That's the entire product. Multiply by every message you send in a week.

What I'd tell past-me

Those years of copy-pasting weren't wasted — they were research. By the time I sat down to build, I had a specific picture of what the tool needed to be:

1. Inline beats in-app. A separate writing tool means a context switch. Whatever the interface looks like, it has to live where the writing already happens.
2. The default tone is the product. Most users will press the shortcut once and accept whatever comes back. If the default isn't opinionated, the product isn't useful.
3. Rewrite, don't polish. The ESL fingerprint is in the sentence shape, not the word choice. A tool that only fixes word choice is a tool that defends the fingerprint.
4. Translate is a feature, not a category. The job is "express this thought as a native would" — the input language doesn't matter.

I wrote this post because someone asked me at a meetup, "why did you build another writing tool — isn't that crowded?" The answer is: the crowded part is grammar. The empty part is dramatic rewrite, inline, for people who already speak the language. That's the gap I was sitting in for years.

If you're in the same gap — speaking fine, writing tentatively, copy-pasting into ChatGPT all day — try the extension. If you have feedback, my email is in the footer.

Related: 12 ESL patterns native speakers rarely use covers the specific phrases the Native tone strips. English writing tips for non-native speakers is the higher-level version.