This weekend, a post by Matt Webb kept appearing in my feeds, shared by Simon Willison and others. It discusses something Matt calls “context plumbing” in AI systems.
This context is not always where the AI runs (and the AI runs as close as possible to the point of user intent).
So the job of making an agent run really well is to move the context to where it needs to be.
But I think we’ve got this backwards.
The real question: Where should AI live?
Instead of moving context to the AI, we should move the AI to where the content already lives. This isn’t just a technical distinction. It’s a fundamental difference in how we think about AI’s role in our work.
The big AI companies want to create “everything apps.” They’re building platforms that pull all your content, all your context, into their systems. And the motivation is clear: they’ve made enormous investments and need to avoid AI becoming a commodity with razor-thin margins.
But what actually benefits you as a user?
AI as feature, not platform
Having AI incorporated into the tools you already use makes more sense than abandoning those tools for an AI platform. Apple may be lagging behind in the AI race by some measures, but their vision resonates with how people actually work. Apple Intelligence in the Mail app is one of my most used AI features precisely because it’s where I already spend time managing email.
Or consider my workflow in Tana, where I manage my work and knowledge. I can run AI, any model I choose, right on my task list. On my meeting notes and transcriptions. In my journal entries. The AI comes to where my thinking already happens, rather than forcing me to export everything to a separate AI platform, or, as Matt Webb calls it, create a plumbing system to get my context into the AI.
This is the difference between AI as infrastructure and AI as destination.
The platform trap
When AI becomes a platform, you face a choice: either duplicate your work across systems or abandon your existing tools entirely. Neither option serves you well. Duplication creates synchronization headaches and version conflicts. Abandonment means losing the specialized features and workflows you’ve built over time.
The alternative is simpler. AI becomes a feature that enhances the tools you already trust. It’s there when you need it, invisible when you don’t, and it never asks you to restructure your entire digital life around it.
I see AI as a feature and have no intention to use it as a platform. The question isn’t whether AI is powerful or useful. It clearly is. The question is whether that power serves your workflow or disrupts it. Whether it meets you where you work or demands you come to it.
The answer shapes not just which tools you choose, but how you think about the relationship between your work and the technology that supports it.
