CHAPTERS
- 0:00 – 0:56
Cold open: the brain as a computer and restoring vision as proof
Max frames the core thesis that the brain can be treated as a computational system connected to the world through a limited set of “wires” (nerves). He previews retinal prosthetics as a concrete demonstration: generating visual percepts in blind patients and iterating via engineering to improve fidelity.
- •Brain-as-computer framing and “brain in a vat” analogy
- •Senses and motor channels as key I/O wires to the brain
- •Retinal prosthesis positioned as a direct interface to vision
- •Early limitations (grayscale, color) and an engineering roadmap to improve
- 0:56 – 2:21
Max Hodak and Science: mission, BCI-adjacent medical devices, and PRIMA
Sarah introduces Max Hodak and Science (formerly of Neuralink) and sets the agenda: PRIMA’s regulatory progress and broader ambitions like sustaining the human experience. Max describes Science as a medical device company aiming for large effect sizes through working with the brain.
- •Science’s mission: apply understanding of the brain to improve outcomes
- •Retinal prosthesis likened to a cochlear implant for vision
- •Focus on high-impact, device-driven interventions
- •Positioning PRIMA as the flagship product
- 2:21 – 2:53
How PRIMA works: implant + glasses + retinal stimulation pathway
Max explains PRIMA’s form factor and mechanism: a sub-retinal chip paired with glasses that project an image onto the implant. The system bypasses dead photoreceptors and stimulates retinal circuitry to send a usable signal to the brain.
- •Target patients: blindness from loss of rods/cones (e.g., AMD)
- •Sub-retinal implant converts projected images into stimulation
- •External glasses with laser projector drive the implant
- •Clinical expansion planned to RP, Stargardt’s, and other diseases
- 2:53 – 4:34
Choosing the interface site: retina vs thalamus vs visual cortex (and early explorations)
Max walks through the design space for restoring vision by selecting where to interface: retina, LGN in the thalamus, or primary visual cortex. He explains why the retina is compelling when the optic nerve is intact, and notes Science explored multiple modalities before committing.
- •Three candidate sites: retina, LGN, and V1
- •Retina favored when optic nerve remains functional
- •Early exploration: gene therapy, electrical stimulation, ultrasound
- •Emphasis on scientific/technical tradeoffs driving form factor
- 4:34 – 6:21
Acquiring Pixium and reaching commercial readiness in Europe (CE mark)
Max recounts identifying the leading approach in retinal electrical stimulation and ultimately acquiring Pixium, which had advanced clinical progress. He then details the milestone of receiving CE-mark marketing approval for PRIMA, enabling commercial sales in Europe.
- •Pixium identified as “state of the art” in late 2022
- •Technology origins: inventor at Stanford, licensed to Pixium
- •Post-acquisition work to reach approvable commercial state
- •July CE-mark approval and near-term first commercial sales
- 6:21 – 9:09
Timelines, engineering cost, and why device iteration can beat drug discovery uncertainty
Sarah probes how Science evaluates feasibility, risk, and cost across programs. Max contrasts device-led brain interfaces with the high uncertainty of drugs: devices offer faster feedback loops and clearer paths to improvement, as seen in cochlear implants and DBS.
- •Three-pillar pipeline: vision, biohybrid interfaces, and perfusion (Vessel)
- •10–15 year view for a broad medical revolution if successful
- •Drug discovery framed as slower, higher “turn over a card” risk
- •Engineering iteration enables compounding improvements (FOV, grayscale, color)
- 9:09 – 9:40
Clinical trial outcomes: what patients could do and what’s next to improve
Max highlights the most important trial result: an existence proof that coherent “form vision” can be restored. He cites patient capabilities like puzzles and reading, while acknowledging current constraints and the roadmap to expand capability and reliability.
- •Existence proof: restored visual images in the mind’s eye
- •Examples: Sudoku/crosswords, reading, scanning text
- •Perceived as “too good to be true” even among observers
- •Current limitations create a clear engineering improvement target
- 9:40 – 12:09
Clinician and public reactions: “the brain is a computer,” definitions of BCI, and tech vs biotech culture
Max addresses why calling the brain a computer provokes pushback and frames it as definitional “bike-shedding.” He also discusses whether a retinal prosthesis counts as a BCI and how Science’s device-first approach aligns more with tech culture than traditional biotech.
- •Why “brain as computer” triggers defensiveness and debate
- •Computers as state machines; no special role for transistors
- •Retinal prosthesis as BCI expands what counts as BCI medicine
- •Funding/culture: more tech-investor-oriented than biotech
- 12:09 – 15:45
The BCI landscape: from “hand substitutes” to rewriting the boundary of the brain
Max lays out BCIs as a broad category, arguing it’s as diverse as “pharma,” not a single bet. He distinguishes low-level control/communication tools from deeper interfaces that add new structure or sensory/motor capabilities—moving toward transformative changes in what the brain can do.
- •BCI category spans many distinct hypotheses and product types
- •Examples: silent speech devices (EEG or other sensing methods)
- •Skepticism about “think-to-order an Uber” due to ambiguity costs
- •Key threshold: from communicating with tools to redrawing the brain’s boundary
- 15:45 – 17:34
Identity, continuity, and uploading thought experiments: what counts as “you”
The conversation shifts to personal identity: why a perfect software copy may not satisfy the desire for survival. Max and Sarah explore continuity (vs replication) and why breaks in experience—like anesthesia—feel different from making a separate duplicate.
- •Uploading scenario: non-destructive scan creates a replica—does that help?
- •Continuity of experience as central to identity and satisfaction
- •Anesthesia as an intuitive case of discontinuity to explain
- •“Creation/annihilation” framing of minds as distinct operators
- 17:34 – 19:49
Studying consciousness: binding of sensory experience and the partitioning problem
Max describes consciousness as parallel sensory streams that the brain binds into one coherent moment, while excluding other combinations. He argues the key scientific challenge is explaining how experiences remain partitioned to an individual brain—why you never get “my vision with your hearing.”
- •Conscious moments involve simultaneous multi-sensory construction
- •Binding problem: why senses are experienced together as one scene
- •Partitioning: why experiences don’t mix across individuals
- •Continuity over time allows identity drift but resists discontinuity
- 19:49 – 22:03
AI and neuroscience converge: platonic representations, alignment, and what’s controversial
Max argues AI models exhibit internal geometries reminiscent of neural representations, supporting the “platonic representation hypothesis.” He claims practical alignment between animal neural recordings and model representations provides evidence, while noting open questions about how global vs local the structure is.
- •Platonic representation hypothesis as a driver of new BCI interest
- •Similar representational geometry in models and brains
- •Practical alignments: neural recordings mapped to model internals
- •Controversy from incomplete understanding (global vs local structure)
- 22:03 – 22:49
What’s most fertile now: doing “neuroscience on models” and building durable businesses
Max suggests that if shared representations are real, AI research becomes a powerful route to studying neuroscience because experiments are faster and more controllable. The discussion then turns to commercialization: building profitable, durable companies with real revenue rather than only long-horizon bets.
- •Neuroscience insights can be easier to obtain from AI models than biology
- •Friends in AI labs jokingly describe their work as neuroscience
- •Commercialization priority: build businesses that can “do this forever”
- •Scaling from first product toward multiple $100M+ run-rate programs
- 22:49 – 25:17
Market, pricing precedents, and scaling PRIMA to larger populations
Max outlines why restoring vision can be a strong business: large prevalence in age-related macular degeneration and high willingness to pay. He references pricing precedents from older retinal prostheses and gene therapies, and describes how next-generation PRIMA expands the addressable market from hundreds of thousands to millions.
- •AMD prevalence: large and growing with age demographics
- •Pricing anchors: prior retinal prosthesis (~$150k) and gene therapy (~$500k/eye)
- •Current PRIMA TAM: hundreds of thousands (US+Europe)
- •Next version aims to broaden eligibility and scale to millions
- 25:17 – 29:00
Beyond vision: perfusion medicine, preserving the brain, and adapting humans for space
Max reframes medicine around preserving what matters most: the brain, which can’t be transplanted without losing the self. He connects BCIs (sensory/motor I/O) and perfusion medicine to improving healthspan and reducing fragility, extending to long-term goals like adaptation for space and substrate independence.
- •Brain as central organ; other organs as “support characters”
- •BCI focus on sensory/motor I/O as a foundational capability
- •Targets: reduce human fragility; address cardiovascular disease and brain-metastatic cancer
- •Preservation and adaptation converge on swap-able parts and substrate independence
- 29:00 – 31:38
Why Science isn’t building a “brain keyboard”: thinking, language bottlenecks, and product focus
Sarah asks why high-bandwidth thought-to-AI interfaces aren’t Science’s main focus. Max argues that speaking/writing is part of thinking, that the brain has an apparent ~10 bits/sec cognitive bottleneck, and that even if brain-typing is useful, it’s a different product category from generating senses or achieving substrate independence.
- •Claim: talking/writing is thinking; “fully formed” thoughts can be illusory
- •Cognitive bottleneck estimate around ~10 bits/sec from multiple lines of evidence
- •Possible future for wearable monologue-to-AI, but not their focus
- •Science prioritizes sensory generation and deeper boundary-changing interfaces
