Local-first technology

Local-first technology begins with a simple question: what should keep working if the internet disappears? In many conservation and community data projects, connectivity is slow, expensive, intermittent, or politically fragile. If the database, model, or payment dashboard only works through a distant service, then the community does not fully hold the tool.

Local-first does not mean anti-cloud. It means the first copy, first decision, and first right to inspect should sit as close as possible to the people and places the data comes from. A phone can collect observations offline. A laptop can store recordings until review. A small server can hold a shared archive. A self-hosted or small AI model can answer some questions without sending private prompts, exact locations, or cultural knowledge to a remote provider.

This is also a lesson in frugal technology. The best tool is often the smallest one that can do the job: less bandwidth, less power, less maintenance, fewer hidden subscriptions, and fewer people needed to keep it alive. Frugal does not mean weak. It means repairable, teachable, portable, and honest about the real conditions of the field.

Self-hosted AI belongs in this framework because AI is not only a model; it is an infrastructure relationship. A community may choose a powerful remote model for translation, summarisation, or analysis when the benefit is clear and the data boundary allows it. But sensitive archives, land locations, sacred stories, identity documents, and refusal decisions may need a local model, a local index, or no AI at all.

The sovereignty test is practical: can the community read its data, run the core workflow, export the archive, audit what was sent out, and stop the system without losing everything? If yes, the technology is closer to soil. It can be tended, repaired, and governed in place.