I read Karpathy’s Digital Hygiene recently. I read it at the time when I was planning to create a simple “links” page—a hub connecting my Goodreads, Letterboxd, maybe even Spotify. A neat index of everything I’m into. However, reflecting on Karpathy’s emphasis on privacy and security, I became increasingly uneasy.

Karpathy underscores the importance of safeguarding personal data to prevent unauthorized access and potential misuse. He advocates for practices like using password managers, hardware security keys, and even a virtual mail service that scans your physical mail and sends it to your email. All to protect one’s digital footprint.

While his focus is on preventing external breaches, I realized that the very act of consolidating and sharing our personal interests online contributes to a different kind of vulnerability: predictability. Each movie rating, book highlight, or playlist becomes a data point, feeding algorithms that can study, imitate, and even anticipate our preferences.

Consider the case of MattGPT, where a laid-off Amazon ex-colleague turned his personal data into a chat interface capable of answering questions about him, including his movie tastes. It’s a clever and brilliant innovation, but it also highlights how a comprehensive digital footprint can be used to create an accurate facsimile of one’s mind and preferences.

Your clone is in Beta

If you post enough online, someone could probably generate a pretty accurate version of you. Not theoretically—literally. Scrape your stuff, train a model, and boom: a decent facsimile of your mind, your tone, your taste.

It’s not even about privacy at that point. I agree with Kierkegaard in that you can’t live authentically without putting something real out there. But now it’s about predictability. About being flattened into a pattern that can be parsed and replayed.

In a world where language models are good enough to write essays, mimic tone, and even predict preferences before you’re conscious of them, the more content you produce publicly, the more easily your psychological blueprint can be reconstructed. Not just by malicious actors—but by algorithms, employers, governments, or even future AI versions of yourself.

Determinism?

This erosion of surprise in our public presence doesn’t just impact privacy—it nudges us toward a deterministic view of ourselves. Every review, post, or rating becomes another data point feeding a predictive engine that seems to confirm we’re nothing more than the sum of our inputs. Once a model can forecast your next move or your next thought, it challenges the notion of freedom. Are your decisions truly driven by choice, or are they simply the fulfillment of probabilities?

We’re not there yet, but what if we were? Maybe I’m overthinking it—but the possibility that a future model could simulate not just your tastes but your decisions, your tone, your worldview… it’s not that far-fetched. It’s sci-fi with a foot in the present. Even if it never becomes perfectly accurate, the trajectory is clear enough to spark some healthy discomfort. And that discomfort has real, practical implications for how we design our digital lives today.

Final Thought

Your inner life has value. In the machine age, it’s monetizable—and cloneable. If someone can predict you, they can monetize you, manipulate you, or worse: simulate you convincingly.

Once your soul becomes a data product, there’s no need to listen to you.
You’ve already been parsed.