Translator · The profession that carries meaning between languages — from the Septuagint and Jerome's Vulgate to Toledo, Nuremberg's booths and the machine-translation age.
The public image of the skill is knowing a foreign language. The professional reality is the opposite: reading the source language deeply is the entry ticket, and the craft that separates translators is writing — producing native prose so natural that no reader senses another language underneath, while every fact, number and nuance of the original survives intact.
The rest of the job is disciplined suspicion. Good translators assume they have misread something, verify terminology against real usage rather than dictionaries, and revise in separate passes for accuracy and for style. The tools have changed — translation memories, termbases, machine drafts — but the core loop of draft, check, cool, re-read has been stable for centuries.
The scarce skill: producing publishable native prose in the right register, from patent claims to novel dialogue, with no trace of the source's syntax showing through.
Reading at near-native depth — irony, dialect, legal boilerplate, wordplay — and knowing precisely when you have not understood, which is the harder half.
Deciding what a reference, joke or politeness level is doing in the source culture and what will do the same job in the target one — the layer machines most visibly lack.
Hunting a term through parallel texts, standards documents and corpora until the real usage is confirmed, then keeping it consistent across a hundred thousand words.
Enough law, medicine, engineering or finance to catch the errors that matter — in technical translation the expensive mistakes are subject mistakes, not grammar mistakes.
Translation memories, termbases, QA checkers and machine-translation post-editing workflows — the industrial layer that now determines a commercial translator's throughput and rates.
Overnight mail from clients three time zones ahead: quoting new jobs by word count and deadline, negotiating rates, triaging what is possible. Freelancers win or lose the week's work in these two hours.
The core session while concentration is highest: drafting new translation at a professional pace — commonly around 1,500–2,500 words a day of finished commercial text, far less for literary prose — with terminology queries piling up in the margins.
A real pause, ideally out of the house. Translation is continuous micro-decision-making, and revision quality in the afternoon depends on actually stopping — a discipline experienced freelancers guard jealously.
Checking the morning's draft against the source sentence by sentence, then a separate monolingual pass for style; then a second, smaller job or machine-translation post-editing under a different set of expectations and rates.
Updating termbases and translation memories, sending the day's client queries in one batch, invoicing delivered jobs, and the unpaid administration — chasing late payers — that comes with a profession that is mostly self-employed.
Evenings keep the source language alive — novels, news, film without subtitles. Deadlines occasionally invade the night before a big delivery, but unlike hospital professions the crunch is the exception, and the commute is a staircase.
Craft knowledge practitioners actually pass on — not motivation.
Defending his Bible work in a letter of 395/396 CE, Jerome laid down the profession's founding rule: "non verbum e verbo, sed sensum de sensu" — render the meaning, not the words. Sixteen centuries later it is still the first lesson of every translation program, and the first thing clients who count words in the source and target must have explained to them.
Luther's method for his German Bible: do not ask Latin how to say it — "ask the mother in the house, the children in the street, the common man in the market," and watch their mouths. The technique — test every rendering against how the target audience actually speaks — remains the working definition of register control.
Schleiermacher's 1813 Berlin lecture posed the choice every translation still makes: bring the author toward the reader (smooth, domesticated) or take the reader toward the author (foreign texture preserved). Lawrence Venuti named the poles domestication and foreignization in 1995. The craft point: choose deliberately per project, and never drift between the two mid-text.
The UN's language services and most international bodies apply the rule strictly: translators work into their language of strongest command, whatever else they read fluently. The reasoning is asymmetric skill — near-native reading is achievable in a decade, near-native writing rarely in a lifetime — and the errors of L2 translation are exactly the fluent-sounding kind that no spellchecker flags.
Literary translators' standard revision technique: the ear catches translationese — unnatural rhythm, calqued syntax, dialogue no human would speak — that the eye, primed by the source, reads straight past. Edith Grossman, translator of Don Quixote and García Márquez, described the whole craft as an act of listening for the author's voice until it can be spoken in English.
Professional revision doctrine separates the passes: first a bilingual pass, sentence against sentence, hunting omissions and shifted meaning; then a monolingual pass with the source closed, editing the translation purely as a piece of native writing — ideally after it has cooled overnight. Mixing the two passes is how errors of both kinds survive.
Computer-assisted translation software — the market was shaped by Trados, founded in Stuttgart in 1984 — splits texts into segments and recalls every sentence the translator has ever translated, enforcing consistency and making repeat work faster. Fluency in at least one major CAT tool is a hiring requirement across the industry.
Curated term collections with definitions, sources and forbidden synonyms. The largest public one, the EU's IATE, holds millions of entries across 24 languages; professionals maintain private termbases per client, because using a competitor's product name once is enough to lose the account.
Searchable bodies of real text — parallel corpora of existing translations, tools like Sketch Engine or Linguee — used to check how a phrase actually behaves in the wild. The professional reflex is to trust attested usage over any dictionary equivalent.
Neural systems — Google's from 2016, Cologne-based DeepL from 2017 — now produce the first draft for much commercial work, consumed inside CAT tools under post-editing workflows standardized by ISO 18587. The translator's added value shifts to catching the fluent-sounding errors machines specialize in.
Monolingual dictionaries and style guides in the target language — the OED, Duden, national style manuals — plus specialist glossaries of law, medicine or engineering. Unglamorous, but the mother-tongue principle makes the native-language shelf the one professionals actually wear out.
False friends and calques produce text that reads smoothly and says the wrong thing — the failure mode machines and tired humans share. HSBC learned the price in 2009 when its "Assume Nothing" slogan surfaced abroad as "Do Nothing," triggering a rebrand reported to cost around $10 million. The craft's defense is systematic suspicion of anything that came too easily.
Falling per-word rates in the bulk market tempt freelancers to compensate with volume — more words, faster, at lower quality, for clients who churn. Machine-translation price pressure has steepened the treadmill. Translators who survive it specialize upward toward liability, voice and expertise; those who race the machine on price lose.
Translating out of your native language, or inside a subject you half-know, produces confident text with buried errors. The cautionary tale is medical: in a 1980 Florida case, Spanish "intoxicado" (poisoned) was taken as "intoxicated," contributing to a misdiagnosis that left teenager Willie Ramirez quadriplegic and to a settlement of about $71 million — cited ever since in training for the language professions.
The scholar paid to profess a subject in public — twenty-five centuries of teaching and discovery, from Plato's grove to the tenure track and the AI tutor.
AI-resistant 76 🏫The world's largest profession: from Sumerian tablet houses to AI-era classrooms, one adult tasked with turning thirty strangers into a room that learns.
AI-resistant 78 📜The trained interrogator of the past — from Herodotus and Sima Qian to the digital archive, turning fragile documents into checkable accounts of what happened.
AI-resistant 70 🧩The scientist of mind and behavior, from Wundt's 1879 Leipzig laboratory to the therapy room — trained to hear what a person cannot yet say themselves.
AI-resistant 82 📖The keeper and guide of recorded knowledge, from Nineveh's clay tablets to the e-lending fight — organizing what humanity knows so that anyone can find it.
AI-resistant 55