2
5 Comments

If you work in a second language, the best writing workflow is to not write in it first

The standard advice for working in a language that is not your first is to compose in it. Practise, immerse, struggle through. For learning that is correct. For getting your job done today it quietly costs you something nobody names.

Here is what actually happens when you draft in your second language. You do not write the sentence you mean. You write the sentence you are confident you can build without error. You reach for the vocabulary you are sure of rather than the vocabulary that is right. You hedge, because hedging is grammatically safer than committing. The output is simpler and more tentative than your actual thinking, and the reader has no way to know that, so they read the simplicity as you.

That is the real tax. Not the extra minutes, the compression of who you are into what you can safely say.

The workflow that fixes it inverts the order. Write the message in the language you think in, at full speed, exactly as precise and as blunt as you actually want to be. Then polish it. Then translate it. You never trade meaning for safety, because at no point are you composing under a constraint.

The result reads differently, and I mean measurably differently in the responses I get. A translation of a precise thought survives as a precise thought. A sentence you rationed while writing it stays rationed forever.

Two honest caveats. First, this is worse for language learning. You practise less, you stay in your comfort zone, and if your goal is to improve you should ignore all of the above and keep struggling in the target language. Second, it only works if polish and translate are close to free. If each message means a round trip to a browser tab, you will do it for the important emails and go back to writing tired English for the other forty.

I build a tool in this space, which is either why I noticed the pattern or why I am biased about it, so take that as disclosed rather than hidden.

The question I actually want answered: if you work in a second language, do you compose in it or translate into it? And did you ever decide that deliberately, or did it just settle that way?

on September 5, 2026
  1. 1

    The compression point you describe is interesting. Have users actually seen better response or conversion rates after switching to “think first, translate second,” or is the benefit mainly perceived writing quality?

    1. 1

      No, and I should correct myself, because I wrote "measurably differently" in the post and that word was doing work it had not earned. I have not measured anything. What I have is my own impression of my own inbox, which is the weakest possible evidence and exactly the kind of claim I spend most of my time telling other people not to make.

      What an actual test would look like, if anyone wants to run it: same person, same kind of message, alternating between composed-in-the-second-language and composed-then-translated, over a few weeks. Measure reply rate, time to first reply, and whether the reply engages with the substance or asks a clarifying question. That last one is the interesting signal, because the failure mode I am describing is not being misunderstood, it is being understood as vaguer than you meant.

      The confound is obvious and probably fatal: you cannot blind yourself to which method you used, and knowing you sent the better version changes how you read the response.

      So the honest position is that I have a mechanism that sounds right and no evidence it produces different outcomes. If anyone has actually tracked this, I would rather be corrected than agreed with.

      1. 1

        That’s a useful distinction. I’d be interested in seeing what the real-world evidence shows. If you’re open to it, what’s the best email to reach you on?

        1. 1

          contact@wrenio.studio goes straight to me, there is no team behind it.

          And if I do run the alternating-method version properly, I will post the outcome in this thread whichever way it points. My honest expectation is that the effect shrinks once it is measured. Most of mine do.

          1. 1

            Thanks! I’ve just sent it over.

            Looking forward to hearing your thoughts whenever you have a chance.