Extensible Software in the Age of LLMs

A short post on Simon Willison's weblog features a quotation from Jeremy Morrell about a new opportunity for extensible software on the web. The core idea is that applications can be built with a solid, accountable foundation while allowing users to safely extend them in numerous directions.

This approach becomes viable because large language models radically lower the cost of authoring extensions. At the same time, modern sandbox primitives reduce deployment costs and establish reliable security boundaries.

By combining these technologies, applications can use large language models to fill in missing pieces and grant users powerful customization capabilities.