On 19th August 2026, Simon Willison posted a quotation from Jeremy Morrell about “Extensible Software in the age of LLMs,” arguing that web applications have a new opening to let users extend software safely. Morrell’s claim was that LLMs reduce the cost of writing extensions, while modern sandbox primitives (isolated execution boundaries) reduce deployment cost and provide security boundaries.

Morrell said developers can build an app around “a solid, accountable core” and then let users extend it “in many directions” by using LLMs to fill in missing pieces. The quotation framed extensible software as a way to give users “super powers,” with LLMs handling extension authoring and sandboxing handling safer execution. Willison labeled the post with sandboxing, ai, generative-ai, and llms.

On 19th August 2026, Simon Willison posted a quotation from Jeremy Morrell about “Extensible Software in the age of LLMs,” arguing that web applications have a new opening to let users extend software safely. Morrell’s claim was that LLMs reduce the cost of writing extensions, while modern sandbox primitives (isolated execution boundaries) reduce deployment cost and provide security boundaries.

Morrell said developers can build an app around “a solid, accountable core” and then let users extend it “in many directions” by using LLMs to fill in missing pieces. The quotation framed extensible software as a way to give users “super powers,” with LLMs handling extension authoring and sandboxing handling safer execution. Willison labeled the post with sandboxing, ai, generative-ai, and llms.