An Overseas Remote Dev Job Pays More Than Double a Local One. English Is the Gate.

Last month I was trying to work out exactly how to pass a certain parameter to an AI API. A search in Chinese turned up a dozen results; click into any of them and the content was nearly identical, all copied from the same blog post written a year ago — and the usage that post described had already changed in the new version. In the end I got it from the official English docs plus chewing through one maintainer's reply in a GitHub issue. It cost me about an extra hour.

This kind of thing happens many times a year. After enough of them, I started turning over a question: should I just spend the time and get my English up to "usable"?

By "working on English" I don't mean studying for a vocabulary test. I mean three concrete things: being able to read first-hand English docs and source comments directly, being able to file an issue on an open-source project and go back and forth with the maintainer, and being able to sit in an English meeting with an overseas client or colleague and write requirements in English. For someone already holding down a job, getting to that level is roughly a few hundred hours.

Whether those few hundred hours are worth it — I eventually landed on one test.

Whether English is a must-have for someone who writes code comes down to one thing first: are you taking the overseas route or not.

On that route, three things all have English as their entrance, and there's no path around it.

First, remote jobs. By the 2026 salary figures from arc.dev and jobicy, remote developers in the US average around USD 97k a year; developers in China average around USD 42k — more than double the gap. That gap isn't because the code is twice as good; it's the difference in what the market behind it can pay. To reach that market, the precondition is being able to do async collaboration in English — a clearly written PR description, making yourself understood in docs and issues. This happens every day, and there's no getting around it.

Second, building a product yourself and charging a global audience. The same small tool, aimed only at domestic Chinese users versus listed on a channel that reaches a global audience taking dollar subscriptions — the size of the pool and the willingness to pay differ by an order of magnitude. But the landing page, the docs, the support emails, the community replies are all in English.

Third, a voice in the open-source community and first-hand information. The genuinely new things in tech mostly come out in English first; the Chinese retelling lags half a year and often gets garbled in the copying (that thing at the start is exactly this). If you want it early, or want to actually step in and join the discussion, you have to be able to read and write.

Touch any one of these three, and English is the gate you must pass.

But this gate has another side.

Do the math honestly: if the buyers of your skill — the ones you have now, plus the ones you can see over the next three years — are all inside the Chinese-speaking circle (your company is domestic, your colleagues speak Chinese, the private clients you take on communicate in Chinese, and what you build isn't meant to sell abroad), then getting English up to "usable" has a marginal return close to zero.

The day-to-day "reading" need is mostly covered by translation tools now. Reading docs, reading release notes — let the AI translate a pass, then scan the original for the key spots, and it's enough. Spending a few hundred hours to get to the level of running meetings and arguing, for that need, is putting the cost where it won't be used. On top of that, language decays when you don't use it; work it up, then don't touch it for half a year, and it slides back — you end up paying tuition twice.

By this point the call is actually clean: decide first whether you want the money on the overseas route, then decide whether to seriously work on English. Do it the other way — work it up first, then look for a use — and it usually fizzles out halfway.

Someone will say: AI translation is this good already, that gate will level out sooner or later.

My view is that the "reading" half really is being flattened, and the "writing" and "holding a conversation" half is not.

Reading is one-directional; if it mistranslates you can still check against the original. But being challenged on your approach in an open-source issue, negotiating with an overseas client on the spot about how the requirements change, replying in an async thread with a message the other person gets in one read — these need you to assemble English of the right tone in real time yourself. Put a translator in the middle and your words lose their tone; the other side asks one follow-up and you can't catch it; a few rounds of that and the trust is gone.

Reading docs and taking part in a discussion are two different abilities. Translation tools make the first one almost free, and precisely because of that, the second one is worth more on the overseas route than before. People who can step in and speak directly have always been a minority.

Back to the original question. I no longer agonize over whether to work on English; I agonize over something else first:

How many buyers of my skill are outside the Chinese-speaking circle right now? Over the next three years, what number do I want that to become?

You have to report that number honestly to yourself. If the answer is "zero, and not planning to change it," then just let the translation tools cover you — don't pay tuition for a gate you're never going to walk through. If the answer is "none now, but I want some," then those few hundred hours aren't a cost, they're lead time. And the further out you go, the more the people who can step in directly are worth, and the more expensive it gets to start.