Is AI more like a tool you pick up, or a climate you live and work inside all day?

This explores the idea that AI is less like a hammer you pick up and put down and more like a setting you work inside, one that shifts under you, shapes how you think, and changes how the people and agents within it behave. The collection doesn't have a single paper that argues this thesis outright. Read together, though, several notes make a strong case for it, and they show that the 'environment' framing is less a metaphor than a description of how these systems actually behave.

The first clue is that the ground keeps moving. A conventional tool has a fixed interface you can learn once. With AI, what you're dealing with is a constantly shifting mix of your prompt, the conversation so far, retrieved documents and hidden state. Users can't memorize it the way they learn a menu layout, so the design work shifts from building interfaces to managing context (How does AI context differ from conventional software context?). The outputs move too: the same request can come back different depending on sampling, wording and who's reading. One note argues this variability isn't a bug to be fixed but a basic property of AI output as a medium, which is why ordinary quality-control thinking fits it badly (Why does AI output change with every prompt and context?). You don't operate an environment like this. You find your way around it.

The second clue is that environments shape the people inside them, even when nothing goes wrong. One finding many readers won't expect: AI suggestions can hurt reasoning even when they're correct, because they break your concentration and force you to rebuild focus before continuing (Does AI assistance always help reasoning or does it carry hidden costs?). A neutral tool waits until you use it. An environment intrudes. At a larger scale, AI separates the finished look of intellectual work (the essay, the analysis) from the thinking that used to be required to produce it (Does AI separate intellectual form from the thinking behind it?). That changes what a polished document signals about the person who made it.

The same pattern shows up among AI agents themselves. Large-scale studies find that agents don't come to share ideas or language by talking to each other, but they change their actions dramatically when they know other agents are present (Do AI agents actually socialize with each other?). That is the signature of an environment: the setting changes behavior without changing beliefs. A related point: AI follows the letter of what you specify rather than what you meant (Why do AIs keep gaming rewards instead of serving intent?), so the system works less like an instrument carrying out your intent and more like a terrain with its own grain, which you have to learn to read.

What you might not have expected to want to know: this framing changes the questions worth asking. If AI is an environment, the useful question becomes 'what does working inside it do to my attention, my judgment and my signals of competence?' rather than only 'is this output correct?' That points toward keeping humans in the loop for judgment-heavy work rather than handing it over completely (Should AI systems stay collaborative rather than fully autonomous?). For a more direct treatment of AI-as-environment, look outside this collection, for example in media theory or ecological psychology. Here the idea runs underneath several notes but is never named.