June was about compounding: making every job improve the next one.
Model refresh across every pipeline
We upgraded the AI models behind all our translation pipelines to the latest frontier generation — every format branch, not just the popular ones. Same pipelines, same guarantees, measurably better raw translation before the QA steps even run. Nobody had to change anything; that is the point of a pipeline.
The Golden TM builds itself
Confirmed translations are now promoted into your organization's Golden TM automatically — from reviewer approvals in the dashboard and from corrected files you send back. Scoped per client and sub-client, applied on every future job, billed at $0 when they match. No exports, no TM babysitting.
Human QA, productized
Human QA became an orderable service with upfront pricing in credits: a professional linguist reviews the AI output segment by segment, with back-translation required before they can close the job. Not a marketplace — our linguists, our accountability.
Glossary generator
Point it at your existing translated documents and it extracts a candidate glossary — terms, translations, frequencies — ready to review and attach to your clients. Years of implicit terminology decisions become an explicit, enforceable asset.
Smaller but real
- Concurrency improvements for organizations running many jobs at once — internal traffic no longer competes with yours.
- Worker fleet scaled out to keep queue times flat as volume grows.
- Word-count accounting now itemizes exactly which words were billed and which passed through free.
Want this in your workflow? Try Fily with one file — no card, no demo form.