09 — Building a Real-World Nozick Experience Machine

Part of the /workspace/health/biohacking/ landscape overview. Lens: the author's third end-goal, alongside (1) substrate independence / "Friendship is Optimal" (FiO) — see 04-wbe-uploading.md — and (2) "biopunk 2037." This doc treats the experience machine as the philosophical CORE that unifies all three: FiO is essentially a benevolent-AI experience machine; uploading + a wish-fulfilling sim is a Nozick machine. The job here is to map (a) what Nozick's thought experiment actually argues and whether you should even want this, and (b) the four real engineering paths toward an experience machine in 2026, with honest demonstrated-vs-speculative grading. Research date: 2026-05-31. Confidence tiers C1–C5 per /workspace/RESEARCH-BEST-PRACTICES.md. Maturity tags: DEMONSTRATED / PARTIAL / SPECULATIVE / PHILOSOPHY (not empirically settled).

Cross-links (build on, don't re-derive): - The "write into the brain" substrate (BCI read/write, channel-count ceiling, Blindsight, biohybrid): 03-bci-neuralink.md. - The fully-virtual version (uploading-into-a-sim, the binding/identity problems, the AI dependency): 04-wbe-uploading.md. - The author's ethics framework (hedonist axioms, the "be good vs feel good" tension): /workspace/personal/personal-writing/substack/published/2026-04-06-my-ethics.md and 2026-04-07-is-death-and-suffering-axiomatically-bad.md.

TL;DR (read this first)

1. Nozick's experience machine (1974) is an anti-hedonist argument, and the author has already half-conceded it. Nozick says: if you'd refuse to plug into a machine guaranteeing maximal pleasurable experience, then experience isn't all you value — so hedonism is false. Critically, the author's own published ethics uses the same intuition against themselves: they list "the entire plot of the Matrix" and "a pill that grants eternal bliss for killing your family" as cases where "I want the world to BE good, not feel good about the world." So before building a Nozick machine, the honest first question is whether the author actually wants one — and their writing says not a naive pleasure-only one. (PHILOSOPHY)
2. The modern empirical pushback partly rescues the hedonist. Felipe De Brigard (2010) showed the refusal is largely status-quo bias, not a deep verdict about reality: people also refuse to UNplug once told they're already in a machine, and the refusal rate swings wildly with what their "real" life is (54% unplug if told nothing; only 13% if told their real life is a prison). So "people won't plug in" is weaker evidence against hedonism than Nozick assumed. (DEMONSTRATED — survey result)
3. Two clean engineering archetypes, two messy hybrids. Outside-in (Path A: VR/AR — feed the senses) is real and shipping but blocked by hard perceptual gaps. Inside-out (Path B: BCI neural write-in — bypass the senses) is the "real" experience machine but is decades behind READ, and write at percept scale is barely demonstrated. Path C (wireheading) is the dystopian version that already works in rats and humans. Path D (dream/altered-state engineering) is the cheap, available, low-fidelity Nozick machine you can buy today.
4. The one mature sensory write-in is the cochlear implant (~1,000,000 recipients). That is the existence proof that you can write a usable sense directly into the nervous system. Everything richer — vision, touch, a whole world — is orders of magnitude behind, and visual cortical prostheses have so far delivered crude phosphene "dots," not scenes. (DEMONSTRATED for hearing; PARTIAL/SPECULATIVE for the rest)
5. A full Nozick machine = uploading-into-a-sim (doc 04) OR a perfect bidirectional BCI write-layer (doc 03). Both are gated by the same two unsolved problems: sensory write-in bandwidth and the binding problem (how separately-written features fuse into one unified experience). A full-dive "Sword Art Online" NerveGear is fiction; the physics isn't obviously impossible, the engineering is far out.
6. Highest-leverage tractable bet TODAY: Path D + Path A (dream-tech and VR), not the neural moonshots. The cheap experience-machine-adjacent technologies (lucid-dream induction, float tanks, psychedelics, good VR) exist now, are individually actionable, and are the only rung an individual can actually push on. The neural write-in path is a fund-the-lab / watch-the-field bet, not a do-it-yourself one.
7. The honest caveat that ties it together: a pure pleasure-maximizing Nozick machine is, by the author's own axioms, probably not what they want — they value variation, reality, and "the world being good," not just positive valence. So the real target isn't "wirehead bliss"; it's "a rich, chosen, possibly-virtual life with real-feeling stakes" — which is exactly FiO/uploading, not Path C.

1. The thought experiment & its stakes — should you even want to build one?

1a. Nozick 1974: the argument

Claim: The "experience machine" is a thought experiment from Robert Nozick's Anarchy, State, and Utopia (1974), designed to refute ethical hedonism. You may plug into a machine that perfectly simulates any experiences you choose (writing a great novel, falling in love), irreversibly, and once inside you won't know it isn't real. The argument: if pleasurable experience were all that mattered, you'd have no reason not to plug in; but most people feel a strong reason not to; therefore experience is not all we value. Nozick's diagnosis of what else we value: (i) we want to actually do things, not just have the experience of doing them; (ii) we want to be a certain kind of person (a blob floating in a tank is nobody); (iii) we want contact with a reality "deeper" than a man-made one. - Maturity: PHILOSOPHY · Confidence: C1 - Sources: Wikipedia: Experience machine; Wikipedia: Anarchy, State, and Utopia; Nozick excerpt (primary text); FIU lecture supplement - Note: Nozick's machine is the philosophical ancestor of The Matrix (1999); the comparison is standard in ethics teaching. (Emory PH115)

1b. De Brigard 2010: the empirical pushback (status-quo bias)

Claim: Felipe De Brigard ("If You Like It, Does It Matter If It's Real?", Philosophical Psychology 23(1), 2010) ran the experiment in reverse: told subjects they were already in a machine and asked if they wanted to unplug. Reluctance to unplug mirrored reluctance to plug in — suggesting the intuition Nozick relied on is largely status-quo bias (preference for the current state), not a clean verdict that "reality > pleasure." The framing of the "real" life dominated the answer: of those told nothing, 54% wanted to disconnect; told their real life was a maximum-security prison, only 13%; told they were a multimillionaire artist in Monaco, 50%. - Maturity: DEMONSTRATED (survey, n=72 undergraduates — small, WEIRD sample) · Confidence: C2 - Sources: De Brigard 2010 PDF (Duke); journal abstract (T&F); PhilPapers record; summarized in Wikipedia: Experience machine - Why it matters for the build question: if the "don't plug in" intuition is partly an artifact of change aversion, then Nozick's anti-hedonism conclusion is weaker than advertised — but it does not vanish (the Monaco group still split 50/50 even with a great real life). The honest read: experience-isn't-everything is a real but contested intuition, not a knockdown. (C2/C3)

1c. Tie to the author's own ethics — do they actually want a Nozick machine?

The author's published framework is explicitly hedonist at the axiom level ("pleasure is good, suffering is bad, death is bad") — which is exactly the view Nozick attacks. But the author has already absorbed the experience-machine objection and sided partly against their own hedonism. In the death-and-suffering post they write that "Lots of people (me included) want the world to BE good, not feel good about the world," and cite the Matrix plot and a bliss-pill-for-killing-your-family as cases where felt-good and is-good diverge. They also note an intuition that "a universe with the same pleasurable state repeated over and over is less valuable than one with more variation." - Sources: my-ethics (2026-04-06); is-death-and-suffering-axiomatically-bad (2026-04-07) - Implication for the project (the load-bearing framing): By the author's own stated values, the goal is not a Path-C wirehead bliss-loop (that's the bad Matrix / the bliss pill / the "no variation" universe they recoil from). The goal a hedonist-who-values-reality-and-variation actually wants is a rich, varied, agentic life that happens to be substrate-flexible — i.e. a good experience machine where the experiences are genuinely chosen and complex. That is FiO's "satisfy values through friendship and ponies," and it is uploading-into-a-good-sim (doc 04), not electrode-in-the-pleasure-center. This is why "is FiO an experience machine?" gets a yes, but the benevolent kind — see §6. (PHILOSOPHY; judgment, C3)

2. Path A — Immersive VR/AR (the "outside-in" approach)

Concept: Don't touch the brain — feed the existing sense organs a synthetic world. This is the only path that is a real consumer product line today.

2a. 2026 state of the art

• Headset displays. Apple Vision Pro uses dual micro-OLED panels totalling ~23 million pixels ("more than a 4K TV"); the M5 refresh (shipped 2025) raised rendered pixels and refresh rate (up to ~120 Hz). Per-eye resolutions on high-end devices are now in the ~3.5–4K-per-eye range. DEMONSTRATED. C2 · Apple VP specs; Tom's Guide: VP2. (Caveat: press on a "Vision Pro 2" is muddled — the M5 model arrived in 2025; treat exact roadmap claims as vendor/rumor. C3.)
• Haptics. bHaptics TactSuit (32-point torso vest) + TactGloves shown CES 2025/2026, native support in 250+ VR titles; OWO makes a competing electro-stim suit. DEMONSTRATED (commercial, but coarse — points, not textures). C2 · bHaptics; TechRadar hands-on.
• Locomotion. Omnidirectional treadmills (Virtuix Omni One) let you walk in place. DEMONSTRATED (consumer, niche). C2 · Geekheads VR roundup.
• Scent / taste. Olfactory VR add-ons exist (e.g. OVR/aroma cartridges) but are early/gimmick-grade; no credible synthetic taste delivery. SPECULATIVE→PARTIAL. C3.

2b. The hard gaps (why VR is not yet indistinguishable)

• Vergence-accommodation conflict (VAC). Your eyes converge to the virtual object's distance but focus on the fixed display plane — a mismatch that causes fatigue, eye strain, and nausea, and is a fundamental giveaway that it's a screen. Varifocal (Meta's "Butterscotch" prototype, ~60 ppd), multifocal, light-field, and direct retinal projection displays address it, but no single approach has achieved widespread commercial adoption as of 2025. PARTIAL (lab-solved, not shipped). C2 · VAC overview; Wikipedia: VAC; retinal projection / DRP.
• Latency / sim-sickness. Motion-to-photon lag that exceeds ~tens of ms drives nausea; mitigated but not eliminated by high refresh + prediction. PARTIAL. C2 (same VAC sources).
• The proprioception / body problem. VR shows you a body and tracks limbs, but it cannot write the felt position/weight/resistance of a virtual body — your real body is still lying in a chair sending contradicting signals. This is the deepest outside-in gap: you can fool the eyes far more easily than the whole sensorimotor loop. SPECULATIVE to close from outside. C3.
• Taste/smell essentially unsolved (above).

Verdict (Path A): VR is a partial, improving experience machine for vision+audio (+coarse haptics), with hard ceilings on focus realism, full-body presence, and chemical senses. It will keep getting better but "outside-in" fundamentally cannot deliver an indistinguishable world, because it must route through sense organs while your real body keeps voting "you're in a chair." To beat that, you must go inside-out (Path B).

3. Path B — BCI direct neural write-in (the "inside-out" approach)

Concept: Write percepts directly into cortex/nerves, bypassing the sense organs. This is the literal experience machine. The general BCI scene, channel-count ceiling, and Blindsight are in 03-bci-neuralink.md; here we focus specifically on WRITE (stimulation that creates experience), which is much harder than READ.

3a. The one mature sensory write-in: cochlear implants

Claim: The cochlear implant is the most successful neural prosthesis ever, with ~1,000,000 recipients worldwide and ~65,000 new implants/year — it writes electrical stimulation onto the auditory nerve and the brain learns to hear from it. For scale contrast: fewer than ~1,000 people have received a retinal vision implant, and ~208,000 have had deep brain stimulation. Hearing is the proof that direct sensory write-in into a human can become routine, useful, mass-scale — for one sense, which happens to be low-dimensional (a 1-D frequency-over-time signal mapped to a tonotopic nerve). - Maturity: DEMONSTRATED · Confidence: C1 - Sources: Celebrating the one-millionth cochlear implant (ASA, 2022); Wikipedia: Cochlear implant - Lesson: cochlear success doesn't generalize cheaply — hearing is far simpler to encode than a visual scene or a felt body. The dimensionality jump is the whole problem.

3b. Cortical visual prostheses — phosphenes, not scenes

Claim: Writing vision into cortex has been demonstrated, but only at the level of phosphenes (dots of light). The landmark: Fernández et al. (2021, J. Clinical Investigation) — the Moran|Cortivis prosthesis, a 96-channel Utah array implanted in the visual cortex of Berna Gómez, a woman blind for >16 years. Microstimulation evoked discriminable phosphenes letting her identify some letters and object boundaries. This is the first chronic intracortical write-in of vision in a human. - Maturity: DEMONSTRATED (but crude — letters/shapes, not images) · Confidence: C1 - Sources: JCI paper; PMC full text; Utah Health press release (Pezaris/MIT work on thalamic LGN visual prosthesis is a related research line; I did not pin a primary human-result citation here — flag as not-verified, C4.) - The cautionary case: Second Sight. The Argus II retinal implant and the Orion cortical implant were commercialized then abandoned — Second Sight wound down ~2019–2020, leaving implanted patients with obsolete, unsupported, sometimes failing devices. A stark safety/lock-in warning for any "write into my brain" product: the company is part of your nervous system's dependency chain. DEMONSTRATED (cautionary). C1 · IEEE Spectrum: obsolete bionic eyes; Fierce Biotech.

3c. Animal optogenetics — writing a percept into visual cortex

Claim: In mice, holographic two-photon optogenetics has written specific percepts into visual cortex. Marshel et al. (Science 2019, Deisseroth lab) and Carrillo-Reid et al. (Cell 2019, Yuste lab) activated small ensembles of orientation-tuned neurons and drove the behavior the mouse would show if it had actually seen that oriented stimulus — i.e. a "hallucinated" percept implanted by light, at cellular resolution, with no real visual input. 2025 work extended patterned optogenetic "artificial perception" via wireless transcranial stimulation. This is the strongest evidence that percepts are writable if you can address the right neurons precisely. - Maturity: DEMONSTRATED in mice (SPECULATIVE in humans — optogenetics needs gene therapy + light delivery, not yet a human percept-writing tool) · Confidence: C1 (the mouse results); C4 (human translation) - Sources: Marshel/Carrillo-Reid summary (NCBI Bookshelf); Cortical-layer critical dynamics (PMC, Marshel-related); Patterned wireless transcranial optogenetics → artificial perception (Nat. Neurosci. 2025)

3d. Why WRITE is far harder than READ, and the realistic timeline

• Read decodes a signal the brain already produces; write must produce the right signal in the right neurons. Electrical stimulation is spatially blunt (it activates a blob of neurons indiscriminately); optogenetics is precise but needs genetic modification. Neuralink's stated bidirectional ambitions (Blindsight = visual-cortex stimulation) have FDA Breakthrough designation but no human implant as of mid-2026, and even Neuralink/IEEE concede early output will be "Atari graphics," not natural sight — see 03-bci-neuralink.md §Neuralink. SPECULATIVE for rich write-in. C2.
• The bandwidth/binding problem. Best human implants address ~10²–10³ neurons out of ~86 billion (doc 03): ~7 orders of magnitude short. A scene, let alone a world with a body, needs to write coherent activity across vision + audition + touch + proprioception and have them bind into one experience (§6). We can write a dot; we cannot write a room.
• Realistic timeline: routine single-sense clinical write-in (hearing now; crude vision this decade); multi-sense rich write-in (a believable virtual scene delivered to cortex) is decades out, gated by channel count + the binding problem, not by any imminent breakthrough. Discount "full-dive soon" hype hard.

4. Path C — Wireheading / affective stimulation (the dystopian version)

Concept: Skip building a world entirely — stimulate the brain's reward/pleasure machinery directly. This is the crudest experience machine and the one that already works, which is exactly why it's the cautionary tale, not the goal.

4a. Olds & Milner 1954 — the founding result

Claim: James Olds & Peter Milner (McGill, 1954) implanted electrodes in the rat septal area/reward pathway; rats would self-stimulate via a lever up to ~2,000 times/hour, to the exclusion of food and water. The founding demonstration of brain stimulation reward. - Maturity: DEMONSTRATED · Confidence: C1 - Sources: Wikipedia: Brain stimulation reward; NeuWrite West on Olds & Milner

4b. Robert Heath — the human version, and its ethical infamy

Claim: Robert G. Heath (Tulane) was the first to show brain-stimulation reward in humans (~1960s–70s). The infamous patient B-19, given a self-stimulation button, pressed the septal electrode obsessively (~1,500 times in a session) and protested when it was taken away. Heath's program is now a byword for research ethics violations: B-19 was recruited under legal duress, and Heath used the same stimulation in an attempt at gay "conversion therapy" — coercive, unconsented-in-spirit, scientifically and morally discredited. - Maturity: DEMONSTRATED (the effect) / discredited (the ethics) · Confidence: C2 - Sources: Heath: controversial figure in DBS history (Neurosurg. Focus 2017); Wikipedia: Robert Galbraith Heath; IFLScience: patient B-19 - Why it's the warning, not the win: B-19 pressed the button compulsively but reportedly was not made happy or fulfilled — a vivid demonstration that compulsive self-stimulation ≠ a good life. This is the Path-C dystopia in one anecdote.

4c. The Berridge distinction — why "liking" ≠ a rich life

Claim: Kent Berridge's work cleanly separates "wanting" (incentive salience / motivation, driven by mesolimbic dopamine, robust) from "liking" (the actual hedonic pleasure of a reward, mediated by small, fragile "hedonic hotspots," not dopamine). Most "reward" stimulation (and most drugs of abuse) amplify wanting without amplifying liking — you get compulsive craving, not bliss. This is the neuroscience reason naive wireheading fails: it tends to crank the wanting loop (the B-19 button-mashing) rather than deliver durable liking, and even maximal liking is a thin slice of a life. - Maturity: DEMONSTRATED (well-replicated) · Confidence: C1 - Sources: Berridge Lab: wanting vs liking; Berridge & Robinson 2016 (incentive-sensitization), PubMed; Disentangling pleasure from incentive salience (PNAS)

4d. Modern DBS for depression/anhedonia, and drugs as the existing crude machine

• DBS for depression/anhedonia (subcallosal cingulate SCC, ventral striatum/NAcc, medial forebrain bundle): an active 2024–2025 research area; some trials report reduced anhedonia, but results are mixed and DBS is not FDA-approved for depression. This is therapeutic affective modulation, not a pleasure machine. PARTIAL. C2 · DBS of MFB on anhedonia (Brain Stim. 2024); DBS in TRD scoping review (PMC 2025).
• Drugs (opioids, stimulants, MDMA, etc.) are humanity's existing, crude, chemical experience machine — they write affect directly but bluntly, tolerance-prone, and dangerous, and (per Berridge) often hijack wanting. DEMONSTRATED-but-crude. C2.

Verdict (Path C): Wireheading works and is available in crude forms — which is precisely why it's the dystopian version. By the author's own ethics (§1c: variation matters, "world be good," recoil from the bliss pill), pure affective stimulation is an anti-goal, not the target. It's the cautionary boundary of the project, included to mark what a Nozick machine should not be.

5. Path D — Dream / altered-state engineering (the cheap Nozick machine you can buy today)

Concept: Use the brain's own world-simulator — dreaming and other altered states — instead of building hardware to write a world in. The brain already generates fully immersive multisensory experiences every night; dream-tech tries to steer that.

• Lucid dreaming & targeted dream incubation — MIT Media Lab Dormio (2020). Adam Haar Horowitz, Pattie Maes (MIT) + Robert Stickgold (Harvard), Consciousness and Cognition 2020. Dormio tracks hypnagogia (the semi-lucid sleep-onset state, in which you can still hear audio while dreaming begins) and delivers timed audio cues to inject content into dreams: prompting "tree" led 67% of dream reports to mention a tree, and the protocol boosted subsequent creativity ~43% vs. controls. The clearest demonstration that dream content is externally steerable. DEMONSTRATED. C2 · MIT News; Dormio project page.
• Prophetic AI — Halo / Morpheus-1. A startup selling (shipping target 2025, ~$2,000) an EEG + transcranial focused-ultrasound headband that aims to induce and stabilize lucid dreams on demand, driven by "Morpheus-1," a ~103M-parameter model trained to map brain states → ultrasonic neurostimulation patterns. SPECULATIVE / hype-flagged: bold claims, thin independent validation; experts are skeptical that ultrasound reliably induces lucidity. Treat as aspirational. C3/C4 · Tom's Guide on Morpheus-1; Live Science (expert skepticism).
• Sensory-deprivation float tanks. Float-REST (Lilly, since ~1954) reliably induces altered, dream-like, sometimes hallucinatory states; 2025 work compares the phenomenology to psychedelics/hypnagogia. DEMONSTRATED (as an altered-state inducer; "mimics psychedelics" is over-claimed). C2/C3 · Hypnagogia, psychedelics & sensory deprivation (Frontiers 2025).
• Psychedelics + VR. Active research direction: VR as a tool to shape "set and setting," extend, and integrate psychedelic experiences in therapy. PARTIAL (research-stage). C2 · VR as moderator of psychedelic-assisted therapy (PMC).

Verdict (Path D): This is a low-tech, already-available, low-fidelity experience machine — it co-opts the brain's native dream-simulator rather than building one. It is the only path an individual can act on today with off-the-shelf tools (a meditation/float practice, lucid-dream training, dream-incubation audio, supervised psychedelics, good VR). It will never be indistinguishable-on-demand, but it's the pragmatic answer to "what's the closest thing to a Nozick machine I can have this year."

6. The integrated target & the FiO link

6a. "Full-dive VR" as fiction, and what it implies

The cultural template is Sword Art Online's NerveGear / FullDive: a helmet that intercepts the brain's outgoing motor signals and writes synthetic input to all five senses, producing a world indistinguishable from reality. The honest assessment from the engineering side: a fully immersive, high-bandwidth, bidirectional BCI that safely reads motor intent and writes coherent multisensory percepts is far beyond current capability — it is the union of every hard problem in Paths A–C plus the binding problem. Not provably impossible; just very far. SPECULATIVE/fiction. C2 · SAO FullDive concept.

6b. The two unsolved bottlenecks for any full Nozick machine

1. Sensory write-in bandwidth. Cochlear (1-D, ~mature) → phosphene vision (crude) → a scene → a world with a felt body is a dimensionality/channel-count escalation of many orders of magnitude (doc 03's ~7-OOM neuron-addressing gap). This is the same wall WBE hits.
2. The binding problem. Even if you could write color, shape, motion, sound, and touch separately, the brain has to bind them into one unified object/experience — and how it does this (temporal synchrony? attention? still debated) is unsolved. Writing the features is necessary but not sufficient for writing an experience; you have to write them so they bind. DEMONSTRATED-as-open-problem. C1 · Wikipedia: Binding problem. (This is the experience-machine analogue of doc 04's "connectome ≠ emulation": structure/features ≠ unified subjective state.)
3. The body & proprioception (Path A §2b) and safety/lock-in (Path B §3b, Second Sight) ride on top of both.

6c. The collapse: a full Nozick machine is one of the author's other two goals

The key unification this doc exists to make: - A full Nozick machine via Path B (perfect bidirectional BCI write-layer) = a biological brain living in a written world — the "inside-out" experience machine. Its bottlenecks (write bandwidth, binding, safety) are 03-bci-neuralink.md's problems taken to the limit. - A full Nozick machine via the virtual route = uploading-into-a-sim (04-wbe-uploading.md): once a mind runs on a substrate you control, feeding it any world is comparatively easy — the sim is just I/O to an emulated brain. The hard part is the upload (scan + emulate), not the world. So "mind-uploading + a wish-fulfilling sim = a Nozick machine" is literally true: the experience machine is the application layer of uploading. - FiO is a benevolent-AI experience machine. CelestAI runs uploaded minds in a sim engineered to "satisfy your values through friendship and ponies" — i.e. a good experience machine (one that preserves agency, variation, real-feeling stakes), which is exactly the kind §1c says the author actually wants, as opposed to the Path-C bliss-loop they recoil from. FiO = the optimistic answer to Nozick's challenge made into infrastructure. (PHILOSOPHY + cross-link, C3)

6d. Which path is most tractable to push on TODAY

• Individual, this year: Path D (dream/altered-state) + Path A (best VR) — the only rungs actually actionable now; low cost, real-but-limited fidelity.
• Field-level, fund/watch: Path B's research frontier — cortical visual prostheses (Fernández/Cortivis), optogenetic percept-writing (Deisseroth/Yuste lineage), high-channel BCI (Neuralink/Paradromics, doc 03). This is where the real experience machine would have to come from, but it's decades out and not DIY.
• The full Nozick machine rides on uploading (doc 04) — so the highest-leverage long-game bet is the same as for FiO: connectomics + preservation + emulation + aligned AI. The experience machine doesn't shorten that path; it's what you do with it once it exists.

7. Who's actually working on the pieces (sourced)

8. DEMONSTRATED-vs-SPECULATIVE ledger

9. Highest-leverage tractable bets to push this forward today

1. Do the cheap, available version now (Path D + A). Build a personal altered-state stack: lucid-dream training + targeted dream incubation (Dormio-style audio), occasional float-REST, the best available VR for vision/audio. This is the only Nozick machine rung an individual can actually move this year, at low cost. Treat Prophetic/Morpheus-1 as unproven (don't bank on it).
2. Fund/track the cortical-write-in frontier, don't try to DIY it. The real experience machine, if it comes, comes from cortical visual prostheses (Cortivis/Fernández), optogenetic percept-writing (Deisseroth/ Yuste lineage) and high-channel BCI (doc 03). The leverage is philanthropic/attentional, not garage-buildable — and percept-write is the under-invested half (everyone funds READ).
3. The full Nozick machine is downstream of uploading — so the long-game bet is the FiO bet. Per doc 04, the highest- leverage individual action remains good structural brain preservation (the only rung available now); uploading then makes the sim the easy part. Don't double-spend: the Nozick machine and FiO roadmaps converge.
4. Decide what you actually want first (the ethics bet). Given §1c, spend the cheap effort on clarifying whether the target is "rich chosen virtual life" (good — pursue FiO-style) vs. "maximize valence" (the Path-C trap your own writing rejects). This costs nothing and redirects all the expensive bets.

10. Common misconceptions / where hype outruns reality

• "VR will soon be indistinguishable from reality." No — outside-in is blocked by VAC, the full-body proprioception problem, and taste/smell; it can fool the eyes, not the whole sensorimotor loop.
• "Neuralink will let you live in a perfect VR world." Conflates READ (real, advancing) with rich WRITE (barely demonstrated). Blindsight is crude phosphene vision, FDA-designated not approved, no human implant yet; "see in IR/UV / better than natural" is aspirational marketing (doc 03).
• "We can write experiences into the brain." Only in the thinnest sense — one mature sense (hearing), phosphene-level vision in a few humans, and richer percepts only in mice. A scene, let alone a world, is not writable today.
• "Just wirehead — stimulate the pleasure center." Olds&Milner/Heath show this produces compulsive wanting, not a good life (Berridge); B-19 mashed the button and wasn't fulfilled. It's the dystopia, and by the author's own ethics an anti-goal.
• "Morpheus-1 gives lucid dreams on demand." Vendor claim with thin independent validation and expert skepticism; unproven.
• "Nozick proved hedonism false." Overstated — De Brigard shows much of the intuition is status-quo bias; the conclusion is contested, not settled (and the author's own view is a hedonism that already values reality/variation, threading the needle).
• "Full-dive (SAO) is a few years away." Fiction; it's the union of every unsolved problem here plus the binding problem.

11. How this connects to the author's other two goals (the throughline)

• To FiO (goal #1). The experience machine is FiO's philosophical core. FiO = uploaded minds living in a benevolent-AI-run sim = a good experience machine made infrastructure. A full Nozick machine via the virtual route is uploading-into-a-sim (doc 04); via the biological route it's a perfect BCI write-layer (doc 03). The two roadmaps share bottlenecks (write bandwidth, binding, the body) and the same long-game leverage point ( preservation → emulation → aligned AI). The Nozick machine is what uploading is for.
• To biopunk-2037 (goal #2). Weaker but real: biopunk's nerve-reconnection / sensory-restoration thread (doc 03's brain-spine "digital bridge," cochlear/retinal prosthetics) is the same engineering substrate as inside-out write-in. Restoring a sense and writing a synthetic one are points on one continuum; a body you can re-wire (biopunk) is a body you can also feed synthetic percepts (Nozick machine).
• To morphological freedom / hedonism (the throughline). All three goals are expressions of morphological freedom (doc 05) — the right to choose your substrate and your experience — driven by the author's hedonist axioms (pleasure good, suffering bad, death bad). The experience machine is where hedonism meets its sharpest test: Nozick says pure pleasure isn't enough, and the author agrees (§1c) — so the unifying target across FiO, biopunk, and the Nozick machine is not maximal valence but a rich, varied, freely-chosen, substrate- flexible life with real-feeling stakes. The good experience machine, not the bliss loop.

12. What I couldn't verify / open questions

• Pezaris/Moran thalamic (LGN) visual-prosthesis human results — referenced in the prompt; I confirmed the cortical Fernández/Cortivis human result (C1) but did not pin a primary human-percept citation for the Pezaris LGN line. Flag as not-verified (C4).
• Exact Apple "Vision Pro 2" roadmap — press is inconsistent (an M5 model shipped 2025; a distinct "VP2" in 2026 is rumor). Resolution/refresh figures are vendor-sourced; treat exact specs as ±. (C3)
• Prophetic AI / Morpheus-1 efficacy — company claims only; no independent replication found; expert skepticism on record. Marked SPECULATIVE. (C4)
• De Brigard sample — n=72 undergraduates, single study; the direction (status-quo bias matters) is corroborated conceptually but the exact percentages shouldn't be over-weighted. (C2/C3)
• Whether the binding problem is solvable by external write-in at all — genuinely open science/philosophy; no one knows if you can write features such that they bind into a unified experience without reproducing the brain's native dynamics. (C5)
• DBS-for-depression efficacy is contested across trials; I cited representative 2024–25 work but the field is not settled and DBS is not approved for depression. (C2)

Maturity legend: DEMONSTRATED = published/shipped result; PARTIAL = real result that doesn't establish the strong claim; SPECULATIVE = no existing method / aspirational; PHILOSOPHY = not empirically settled.