![[I am the same I, that I was yesterday - William James.jpg]] [^1] --- Recently, as I sat at a lunch table with some colleagues, we did what people in our world do in 2025/26. Avoided politics, and instead talked about the possible risks and boons of AI on the horizon. While we agreed about many of the likely near-term outcomes, they didn't have a response to a problem I couldn't quite name at the time. ## The Interface Shift If the future of white collar work has us all interacting with a "personalized" work AI rather than traditional interfaces of files, browsers, etc., what will this mean for the individual contributors themselves? The problem revolves around the projection that in the not-so-distant future the way we interact with AI will be less app, more interface. Rather than logging in to your work machine and opening Word or Chrome, and maybe invoking your AI of choice via a toolbar button, we'll just… chat and/or be served tasks and information by AI. This AI agent will know our job, our goals at work, our skillset, and our calendar. Imagine if Copilot was as useful as Clippy was friendly. ## Beyond Surveillance Yes - as [Ethan Mollick](https://en.wikipedia.org/wiki/Ethan_Mollick) and others have pointed out workplace surveillance as we know it may very well get much much worse at many companies, but what I'm wondering about is more than that. While a Clippy-Big Brother hybrid is equal parts horrifying and somehow just about on brand for the timeline that brought both predictive crime police and crocs to fruition, it's made me think beyond "will AI replace X job"? Some jobs will be lost to automation, this much is clear and LLMs in the corporate sphere are already being used to make better "predictions". While [Agentic AI](https://en.wikipedia.org/wiki/Agentic_AI) hasn't taken off with quite the fervour expected, however… Agentic interfaces increase extraction of behavioural data. The agent mediates every action, creating a unified [audit trail](https://en.wikipedia.org/wiki/Audit_trail) and preference model. ## Employability as Legibility What does this do to anyone's employability? From my perspective employability becomes legibility-to-the-system plus access to proprietary context - in other words, you fit the bill for the kind of "connector" a corporation is looking for and they give you access to their information to connect their dots. When the only real measure of a "human in the loop" is to connect information in a way that only "you" can, what is the value of "work" and how do individuals protect their [intellectual property](https://en.wikipedia.org/wiki/Intellectual_property) from a threat so novel? Suddenly something new takes form, not simply AI replacing jobs, but AI redefining what "work" means via a kind of [[capital cognitive capture]]: when your "style," decisions, and feedback become training data and managerial [telemetry](https://en.wikipedia.org/wiki/Telemetry). I want to hold two truths at once without cancelling either. - First: if AI becomes the interface, the interface becomes governance, and governance becomes extraction by default unless we build alternatives. - Second: even perfect tools cannot save a culture that forgets how to relate. ## The Novelty Problem Intellectual property has historically been a "first to print" kind of commodity - beyond patent law and Dragon's Den. The implications this seems to express in our lives however, was poignantly articulated by [Brandon Sanderson](https://en.wikipedia.org/wiki/Brandon_Sanderson) in his fantasy book The Way of Kings a decade ago: > "Is it the intellect of a genius that we revere? If it were their artistry, the beauty of their mind, would we not laud it regardless of whether we'd seen their product before? But we don't. Given two works of artistic majesty, otherwise weighted equally, we will give greater acclaim to the one who did it first. > > It doesn't matter what you create. It matters what you create before anyone else. > > So it's not the beauty itself we admire. It's not the force of intellect. It's not invention, aesthetics, or capacity itself…it must be nothing more than novelty." ## What Becomes Scarce Recently, realizations about the reality of AI training has triggered unresolved (and even yet to begin) disputes about consent, compensation, and copyright. But underneath the legal fights is a more basic question: what kinds of relationships are we building with each other when our tools can copy at scale? We need *sovereign*, *values-encoded infrastructure* because extraction will scale if we let it. And we need a *relational culture* because even "good" infrastructure collapses if we treat one another as opponents, datasets, or disposable commentators. So yes, maybe we should get less obsessed with being first. Not because originality stopped mattering, but because what becomes scarce is trust, informed consent, and the capacity to create together without someone quietly turning the whole relationship into an asset. The hopeful future is not "everything is a remix." It's "we can remix without theft, and collaborate without capture." --- [^1]: Image sourced from the [Public Domain Image Archive / Internet Archive / University of Toronto Libraries](https://pdimagearchive.org/images/b0120127-5615-440d-a809-184c464edc52)