“Through the Binary Lens” brings together articles where code meets culture. By weaving technical and human perspectives, this series uncovers how software, language, and global expansion influence one another in ways both seen and unseen.
We built an app to map the world’s 2,719 endangered languages. Curious how many have disappeared since 1950?
Using UNESCO’s 2010 endangered language data, we built a global atlas and designed a predictive model inspired by Harmon & Loh’s work on linguistic evenness.
One unintuitive result: a language with 50,000 speakers can be closer to extinction than one spoken by 500 families raising children.
At orb, we’ve long pushed back against monocultural globalization. From building niche datasets for AI training (Hausa, Igbo, Yoruba), to language preservation efforts (Judeo-Kashani, Temani), to localizing tools and projects that genuinely serve the public good (Khan Academy, Mozilla, Bitcoin). The uncomfortable truth is that we’re losing ground.
As budgets tighten, what survives is scale: English-first systems, generic messaging, and ad spend optimized for reach rather than relevance. English-first is cheaper, faster, and easier to scale. We’re fortunate that some long-term collaborators, especially in pharma, public-sector manufacturing, and highly technical industries, still value serious L10N, where precision, trust, and nuance are non-negotiable. We’re also watching zero-source content systems mature. If done right, they could keep humans in the loop by design as technology improves and costs fall.
In the meantime, we’ll keep doing what we can: preserving linguistic nuance, defending global expansion as craft, and making sure people can recognize themselves in the language that speaks to them, rather than sacrificing their voice to broken copy and erratic ads.
Consult the atlas here.
