“Many of the translators’ experiences described in [last week’s newsletter] are familiar from my own work. I’m a software engineer. The article frames the challenges faced by translators as ‘opposite’ to those of software engineers. In fact, our jobs have also been made harder; the most creative activities of the job are being replaced by the tedium of checking machine output; and as we write less code, we feel our skills atrophy. Compared to before, the work offers less scope to demonstrate style and originality. We also face downward pressure on career opportunities and rates of pay. It seems that large language models are having similar effects on white-collar labour across different industries and professions.” Joshua Mostafa

I think you are missing a large part of the developing picture. The majority of developers used by large UK and US companies are not in those countries, they are supplied by services firms in near or far shore locations. As the adoption of LLM development grows, several things happen, one of them is development time lines collapse, placing a far higher emphasis on agile practices; sprints that were once measured in weeks are now completed in hours. Co-location or at least time zone aligned and access-able (can get to the office) become far more important. As one commentator mentioned, when you used to have a team of 50 developers but now only need 5, labour arbitrage (near/far shore vs onshore) becomes irrelevant and is overtaken by the amplified needs of greater agility. Based on this I would expect demand for local talent to be robust to strong at the expense of the services companies that are used to supply the near/far shore labour.

I can offer my experience as someone working in a junior-ish capacity at the moment, using Claude daily. The junior-senior hiring bifurcation highlighted in the article makes sense. LLMs are superb at spitting out small chunks of reasonably good code if asked to do a reasonably specific thing. This was what junior developers used to do. My day to day focus is now on “architectural” decisions (I.e. how do we best assemble all of those small chunks of code to do something complicated?) I’m fortunate in that I’m being coached by a senior developer on this. This aspect of development is often overlooked by non-technical leaders because, yes, in the short term, you *can* mash a load of LLM-generated code together and be somewhat confident it’ll do the thing you’ve asked it to do… but you run out of runway quickly as your product becomes more complex. It’s also a non-starter for anything that could end with a lawsuit. That’s why seniors are still very much in demand. TL;DR - Human developers are probably not going away any time soon, but a gap is almost certainly opening up between those who only know how to write code, and those who know what code needs to be written.

“It’s striking how many generalist translators in mid-market segments — jumping from one field to another, working through translation agencies — have been hit by growing use of AI. But that’s not the whole picture. There are also highly specialized markets for expert translators working directly with risk-averse clients in critical areas, [for example] document types and situations where a flawed translation can create serious legal, reputational or financial exposure — and, in some cases, a threat to life and limb. The translation agencies lowering prices and farming out machine-generated translation for post-editing at rock-bottom rates are rarely present in these segments, since their business model is built for volume. But for genuinely specialized translators prepared to head over and mingle with demanding clients, who have serious writing skills and have invested the time to develop deep expertise, prospects are not all that bleak. (A bit like journalists, perhaps?)”

“There is no denying that the translation profession has become more challenging. Once upon a time, translators were the gatekeepers between languages. If you wanted to know what a page of German said, you had to hire a translator. Those days are gone. For basic cross-lingual comprehension, LLMs and DeepL now perform remarkably well. However, while translation tout court is increasingly commoditised, the combination of translation expertise, deep subject-matter knowledge, excellent service and strong writing skills remains very much in demand. I struggle to see that changing anytime soon. Ultimately, clients are not paying for words to be transferred from one language to another; they’re paying for confidence in the end result.”

“AI is not uniformly disrupting writing. It is destabilizing a particular expectation—that writing should transparently reflect an individual voice. Where that expectation is strong, AI feels threatening. Where writing is already understood as the manipulation of language within a system, it feels like a tool…For example, my other language, Arabic, operates with a structural distinction between spoken dialects and Modern Standard Arabic, the language of writing. Writing in Arabic is not a direct transcription of speech; it is a conversion into a formal system with its own grammar and vocabulary. Linguistic and psycholinguistic research reflects this: written Arabic behaves more like a second language in processing—slower, more effortful, and dependent on conscious control. Literacy itself is often described as overcoming the “intrusion” of the spoken vernacular. In that context, writing is unambiguously understood as transformation. The writer does not simply “sound like themselves.” They produce language that meets the requirements of a standard. English contains the same transformation-heavy domains but tends to frame writing, especially in literary and educational contexts, as the expression of an authentic voice. That expectation shapes how AI is perceived.”