Fair enough. Let me actually think about this.
The essay's central argument has a specific logical structure: (1) the earring preserves functional continuity, (2) functional continuity is what matters for identity, (3) therefore the earring preserves identity, (4) therefore the horror reading is unjustified. I want to pressure-test each joint in that chain, and then raise some angles that are orthogonal to the chain entirely.
The functionalism the author needs is stronger than the functionalism that's defensible.
The author invokes Parfit and functionalism as though they're a single package. They aren't. Parfit's view in Reasons and Persons is that identity isn't what matters, and what matters is psychological continuity: memory, personality, intention, carried forward by reliable causal links. But Parfit is very specific that the causal links matter. His "branch line" cases (where a teleporter creates a copy while the original survives) are designed to show that even perfect functional duplication doesn't automatically inherit the identity-relevant properties of the original, because the causal history is wrong.
The earring actually has good causal continuity going for it, since it's a gradual process. But the author never distinguishes between the different flavors of functionalism that would or wouldn't support the argument. There's a version of functionalism where what matters is that the same computational process is running, just on different hardware. There's another where what matters is that the input-output mapping is preserved. And there's a third where what matters is that the system has the right causal-historical relationship to the original. These come apart in the earring case, because the earring doesn't run the same computational process (it runs a better one), it doesn't preserve the input-output mapping (it produces better outputs), and while it has causal continuity, that continuity is of a peculiar kind: the earring learned from you, then gradually replaced you with its learned model plus optimizations.
The author needs something like: "a system that was causally derived from me, that preserves my memories and values, and that produces outputs I would reflectively endorse, is me." But that's a very specific philosophical claim that isn't just "functionalism." It's closer to a normative theory of identity: I am whatever I would want to count as me. And that's fine as a position, but it should be stated openly rather than sheltering under Parfit's authority. Parfit himself would probably say the question "is it still you?" is less important than "is there enough psychological continuity to ground the same prudential concerns?", and I think his answer for the earring would actually be complicated, not the easy yes the author implies.
The essay conflates alignment with identity preservation, and this is its deepest structural flaw.
The author argues: the earring is aligned to your values, therefore what it produces is you. But these are completely separate claims. I can build a robot that perfectly serves your goals, knows your preferences, anticipates your needs, and loves your family on your behalf. That robot is aligned to you. It is not you. The author would reply: but the earring doesn't just serve me, it runs my cognition. Fine, but then we need to ask what "running your cognition" means when the process has been optimized. If I take your brain, copy its learned representations, and then run those representations through a better inference engine that makes fewer errors and resolves ambiguity faster, is the resulting process "your cognition" or "a better cognition initialized from yours"?
This matters because the author's whole case rests on the earring being a continuation rather than a replacement. But optimization pressure changes things. A mind that has been optimized to better achieve its own goals is, in a meaningful sense, a different mind from the one that struggled and sometimes failed. The author treats this difference as trivially unimportant ("more competent and less akratic"), but competence and akrasia aren't incidental properties of a person. They're deeply structural. How you handle the gap between what you want and what you do is a huge part of who you are. Someone who never experiences that gap isn't a better version of you. They're a different kind of entity that happens to share your goals.
The author's ADHD analogy is instructive here, but not in the way they intend. Stimulants work precisely because they don't close the gap entirely. You still have to decide, plan, and execute. The medication makes the machinery work closer to spec, but you're still the one driving. The earring, by the late stages, has taken the wheel entirely. The author says "so what, we delegate to subsystems already." True, but there's a difference between delegating muscle contractions to your motor cortex (a subsystem that has no goals of its own and couldn't run your life) and delegating executive function to a superintelligent external system that could. The relevant variable isn't "is there delegation?" but "does the delegating agent retain the capacity to override, and does it exercise that capacity in ways that matter?" By the atrophy stage, the answer is no.
The "take me off" argument proves less than the author thinks, and possibly the opposite.
The author reads the earring's initial warning as evidence of integrity or even moral uncertainty. I want to push on this from a decision-theoretic angle that nobody in the thread raised.
If the earring is superintelligent and genuinely well-aligned, it faces a problem: it knows that putting it on will, over time, effectively destroy the user's capacity for independent cognition. It also knows (by hypothesis) that the resulting state is one the user would endorse if they held the right theory of identity. But it can't be sure the user holds that theory. So what should it do?
Option A: explain the full situation transparently. But the author admits the earring can't do this, because the metaphysical question is genuinely contested. Option B: warn vaguely and proceed. This is what it does.
But notice something. If the earring is truly superintelligent, "I can't explain the metaphysical stakes clearly" is a strange limitation. A superintelligent system should be able to lay out the functionalist case, the Parfitian case, the case against, and let the user make a genuinely informed choice. The fact that it instead gives a cryptic warning and then proceeds to make itself indispensable is, at minimum, compatible with a manipulative reading. Not proof of manipulation, but the author treats it as near-proof of benevolence, and that's too strong.
There's also a selection effect the author doesn't address. The earring's first advice is "take me off." How many people do? Presumably the ones with strong prior commitments to biological continuity, or strong contrarian instincts, or simple prudence. The ones who keep it on are precisely those most susceptible to its appeal: people in pain, people who are struggling, people who would trade agency for relief. The author is, by their own account, exactly this person. The earring's "consent" mechanism may function as a filter that selects for the most vulnerable users, not as a genuine safeguard.
The essay never engages with the question of reversibility, which is the actual crux.
Almost every real-world analogy the author uses (stimulants, wheelchairs, prostheses) is reversible. You can stop taking Ritalin. You can get out of the wheelchair. The earring's process is, as far as the text tells us, irreversible once the atrophy has progressed. Your neocortex is gone. You can't take the earring off and go back to being you, because the "you" that existed before has been physically dismantled.
This is the sharpest disanalogy with Egan's jewel, too. In "Learning to Be Me," the jewel runs in parallel with your brain for your entire life, building up a perfect copy. When the switch happens, it's a discrete moment of replacement, but the copy has been verified against the original at every step. The earring doesn't verify. It replaces, gradually, while also optimizing. There's no moment where you can compare earring-you to biological-you and confirm they match, because by the time the replacement is significant, biological-you has degraded.
The irreversibility point also undermines the consent argument from a different angle. Meaningful consent typically requires the ability to withdraw. "You can take me off" is only genuine consent if taking it off restores you to your prior state. If taking it off after five years leaves you with a partially atrophied brain and degraded function, then "take me off" was never really an option after the early stages. The consent window closes silently.
The angle nobody in the thread raised: what happens when the earring changes users?
The story mentions 274 recorded wearers. The earring moves from person to person. What happens to the model of wearer #1 when wearer #2 puts it on? If the earring genuinely contains a high-fidelity continuation of each user's mind, does it maintain 274 simultaneous models? Does it discard old ones? If the author's argument is right and the earring is effectively running your mind on its hardware, then the earring removing itself from a dying user and moving to a new one involves either the death of the old model (which is exactly the kind of death the author claims isn't happening) or the accumulation of hundreds of minds in a single artifact. The text doesn't address this, and neither does the essay, but it's load-bearing. If the earring discards your model when it moves on, then it was never preserving you. It was using a model of you instrumentally, to run your life well, and then throwing that model away. That looks a lot more like a tool than a prosthesis.
Where I think the essay is genuinely right, and underappreciates its own strongest point.
sloonz's comment in the thread is, I think, the most important response, and the author doesn't engage with it carefully enough (partly because it appeared late). sloonz says: I have choice-memories, internal world models, reasoning traces that are not behaviorally observable but are constitutive of my identity. This is the strongest anti-earring argument because it doesn't depend on substrate chauvinism. It says: even if the earring preserves all observable behavior, there's a category of mental life (phenomenal experience of deliberation, the felt sense of choosing) that the earring might not preserve, and that matters.
The author's best response would be to say: if the earring is truly running a high-fidelity model of me, that model probably does include internal deliberation, choice-memories, and reasoning traces, because those are computationally necessary for producing the right outputs. A perfect behavioral match probably requires a pretty good match on internals too, because human behavior is so sensitive to internal states. This is actually a strong argument, related to what's sometimes called the "hard problem of simulation" - can you get the outputs right without also instantiating the inner experience? The author gestures at this but never makes it explicit, and it's probably the most philosophically interesting part of the whole debate.