I have a habit of saving screenshots without knowing exactly why. Sometimes it’s a website with an interesting layout, a color combination I might use later, or a piece of writing that caught my attention. Most of it sits around for a while before I figure out whether it’s useful, and quite a lot of it never becomes anything at all.
Until recently, software couldn’t do much with this kind of material beyond storing it and helping you find it again. AI is changing that. An assistant with access to your files, notes, and browsing history can start making connections between things you haven’t organized or even properly thought through yourself.
There’s obvious value in that, although I wonder whether we’ll eventually become more selective about how much of our working lives we make available to AI.
When everything becomes context
In June 2026, Apple announced further Apple Intelligence features across its devices, while Microsoft’s Recall offers an interesting glimpse of how computers might remember our activity. On supported Copilot+ PCs, users can opt in to saving screen snapshots that are processed and stored locally, making it possible to find something they saw earlier without remembering its name or where it was saved.
These features could make computers considerably more useful, especially for people who spend their days moving between documents, browser tabs, and different projects. They also change our relationship with information that used to be difficult for software to interpret.
A collection of screenshots might be the beginning of a design direction, but it might also be a collection of unrelated things that happen to interest you. When I’m still figuring something out, I don’t always want an assistant to identify the pattern, summarize the material, or suggest what I should do next. Part of the work is spending time with those references and gradually developing an opinion of my own.
There are practical reasons to be selective, too. In March 2026, OpenAI discussed the risks of prompt injection, where instructions slipped into documents, emails, or websites can manipulate AI agents into taking unintended actions. Restricting an agent’s access and permissions helps limit the damage if something goes wrong. As AI becomes more capable of acting across applications, deciding what it can reach will become an increasingly important part of using it responsibly.
Choosing what to share
I suspect that by 2027, we’ll start paying more attention to something I’d call intentional context: giving AI the information it needs for a particular task, rather than allowing it access to everything by default.
You might share a specific document when you need a summary, or a piece of code when you’re trying to fix a bug, without connecting your entire workspace. The distinction sounds small, but it gives you more control over how and when AI becomes involved in your work.
It’s also part of the thinking behind Tansei, the little shelf I built for Mac and Windows. I wanted somewhere to keep screenshots, links, files, and bits of text while moving between apps, without having to organize them into a project or start a conversation with an assistant. Tansei keeps those things on your computer, doesn’t require an account, and doesn’t send them to an AI service. You can simply leave something there until you need it.
Tansei
Try Tansei for Mac and Windows
I use AI regularly, and I expect it to become a much bigger part of how we work. But I also think we’ll get better at recognizing when we want its help and when we’d rather work something out ourselves.
We’ve already been through something similar with notifications. For a while, every app wanted our attention, and eventually we started looking for ways to get some of it back. As more software begins offering to understand our work, I wouldn’t be surprised if the ability to leave things uninterpreted becomes equally appealing.
Some ideas need a little time before they’re ready to be explained, even to a very capable computer.
Frequently asked questions
What is intentional context in AI?
Intentional context means deciding what information an AI assistant can access for a specific task. Instead of sharing an entire workspace or history, you provide the files, notes, or details it actually needs.
What is an AI-free workspace?
An AI-free workspace is a place where you can collect and work with information without AI automatically analyzing or acting on it. It doesn’t mean you never use AI. You choose when to involve it.
Is giving AI less access safer?
Limiting an AI agent’s permissions can reduce the damage it might cause if it makes a mistake or encounters malicious instructions. It’s one useful security measure, alongside other protections.
Does Tansei use AI?
No. Tansei is a local-first shelf for Mac and Windows that keeps screenshots, links, files, text, colors, and prompts within reach. It doesn’t use an AI assistant to analyze your items, doesn’t require an account, and keeps your shelf on your computer.
Back to top