Quoting Jeremy Morrell
Simon Willison’s Weblog Simon Willison
Opinion — commentary, not a factual news event.
LLMs could make custom software extensions cheap and safe, Jeremy Morrell says. He thinks that gives apps a new way to let users add powers without wrecking the core.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Jeremy Morrell argues that the web may be entering a fresh phase for extensible software. His bet is simple: large language models have cut the cost of writing extensions, while modern sandboxing has made it cheaper to ship them without tearing up the safety boundaries around the app itself.
That combination, in his view, changes the shape of software design. Instead of asking one team to anticipate every workflow, the app can stay as a solid, accountable core and let users extend it where they need to. The missing pieces get filled in by LLMs.
It is an appealing idea because it treats the model as a builder, not just a chatbot. The system does the careful part; the model handles the custom bits. And if the sandbox is doing its job, those custom bits don’t have to turn into a security nightmare.
Morrell’s line is really a pitch for controlled openness. Give users room to adapt the software, but keep the center stable. In his words, that could mean giving users super powers.
My take — AI-written commentary, not fact-checked reporting
This is the rare AI argument that sounds useful instead of merely expensive. The smart move is not to bolt LLMs onto everything and call it innovation; it is to use them where they reduce the pain of customization without blurring the line between core and chaos. Open enough to be useful, closed enough to survive—that’s the boringly correct take, which is usually the one that ships.
Read more about this at: Simon Willison’s Weblog
