Blog
Notes from the work.
How we think about product engineering, applied AI, and building software that lasts.

Working with translation vendors: tools that keep everyone aligned
Clear assignments, shared queries and traceable feedback keep translation vendors aligned without making email the only record of a decision.

Building a localization pipeline that scales past ten languages
Connect source changes, translation, review and delivery with stable IDs and clear ownership, so each new language follows a dependable process.

How to measure translation quality without arguing about it
Shared categories, severity rules and segment-level evidence make quality scores useful. Start with agreement, then make the arithmetic reproducible.

Human in the loop: where reviewers still beat language AI
Automate the repeatable checks. Give reviewers the context, evidence and authority to handle decisions that depend on meaning, intent and consequence.

What machine translation gets wrong, and how to catch it
Fluent text can still change the meaning. Learn where terminology checks, structural validation and human review catch different translation errors.

Localization quality is a workflow problem, not a translation problem
Quality breaks when context, corrections and decisions get lost between teams. Fix the handoffs from detection to review, arbitration and release.
A little ambition goes a long way
What are you building next?
Bring the idea, the challenge, or the question.
Let’s find the right way forward.
Good conversations make great beginnings.