Jeremy Morrell Outlines the Rise of Extensible Software

In a recently highlighted perspective on software architecture, technologist Jeremy Morrell outlined a shifting paradigm for web applications, pointing to a major opportunity for what he terms extensible software. Morrell argues that the combination of large language models and modern sandboxing technologies has fundamentally changed how developers can approach user-driven customization. By leveraging these tools, software creators can maintain a highly reliable core application while empowering users to build their own custom features on top of it.

According to Morrell, the primary catalyst for this shift is the dramatic reduction in the cost of authoring software extensions. Historically, customizing software required significant programming expertise, limiting extensibility to advanced users or dedicated developers. Large language models change this dynamic by lowering the technical barrier to entry, allowing users to describe desired features in plain language and have the AI generate the necessary code. This effectively democratizes software customization, allowing everyday users to tailor applications to their specific workflows.

However, allowing user-generated code to run within an application introduces significant security and deployment challenges. Morrell notes that modern sandbox primitives solve this problem by lowering deployment costs and establishing robust security boundaries. These isolated environments ensure that user-created extensions can run safely without risking the integrity of the core application or compromising other users' data. This balance allows developers to offer deep customization options without sacrificing the stability of their primary platform.

For software practitioners, this approach redefines the relationship between product development and user feedback. Instead of trying to build every niche feature requested by a diverse user base, product teams can focus on perfecting what Morrell calls a solid, accountable core. Users are then given the tools to safely extend the software in multiple directions, effectively filling in any missing functional pieces themselves. This model not only reduces the development burden on core teams but also provides users with unprecedented agency over the tools they use daily.