Nothing is more annoying and disappointing than a lost opportunity. And for years, the translation industry has been remarkably good at generating them. We had great UI translation tools, until software development methodology changed and left them with no UI left to visualise. CAT tools standardised their exchange formats, but left enough gaps in the specifications that true interoperability never quite arrived. Machine translation entered the market and created the post-editor role, which also handed everyone a convenient excuse to ignore input quality and expect the translator, now rebranded as a post-editor, to work with substandard output and like it. Then the LLM revolution came, and TMS and CAT vendors pretended nothing had changed: they bolted prompt wrappers and chatbots onto the same legacy process, ignoring the fact that sticking to segmentation guts most of the benefit modern AI could offer. A bit harsh? Perhaps. But all of this has gradually turned the industry into a place where everyone quietly follows processes we all secretly know are wrong. We probably can't change that at the industry level. What we can do is rethink the modern translation process from scratch, and apply it wherever possible. The biggest mistake was focusing on corporate process and ignoring the core of the whole business: human translators. And what do translators actually want? Besides money, justice, coffee, and some recognition, they want context, because context directly shapes the quality of their work.

Translate in Context

Contextual translation using multi-layer agentic pipeline

Contextual translation using multi-layer agentic pipeline