This text is a first-person narrative from a designer of advanced retinal implants. The plot centers on a client's request to create a prosthesis that allows her to see her deceased loved ones, not as supernatural apparitions, but as projections generated from her own memories overlaid onto her real-time vision.
The project involves a multi-disciplinary team developing a three-part neuro-prosthetic system: a retinal implant, a cortical bridge, and a hippocampal recall interface. The narrative details the significant technical and ethical hurdles encountered. These include initial hardware failures, the AI's inability to differentiate grief from other negative emotions like fear and trauma, and the instability of projections based on a single individual's memory. Solutions are engineered through advanced AI training ("valence gating") and by augmenting the client's memories with those of friends and family ("auxiliary memories").
While the technology is ultimately successful, the post-implantation phase reveals severe psychological risks. The client initially withdraws from life, becoming consumed by the visions. Although she eventually learns to integrate the technology healthily, the narrator concludes that the potential for psychological harm is too great and resolves not to undertake such a project again. As far as I can tell, the story is a cautionary tale about the unintended consequences of translating emotional states into sensory data.
Building a Thanatopic Eye: A Step-by-Step Breakdown
Step 01. Initial Request & Technical Premise: The narrator, a designer of bespoke retinal implants, receives an unusual request from a client, April: she wants to see her dead husband. The core challenge is not supernatural but technical: translating her memories into visual data and integrating it with her live optical field.
Step 02. System Design: The narrator collaborates with Aux, an expert in "imagination tech." They devise a 3-part system: a retinal implant (screen), a cortical bridge (projector), and a hippocampal recall interface (playback). The initial data gathering involves recording April's neuron-level brain activity as she recalls memories, aiming to train an AI on her subjective experience.
Step 03. System Failure & Ethical Concerns: The team builds and tests the components. After fixing a firmware bug in the cortical bridge with a consultant named Lyre, they test the integrated system on a "phantom cortex." The test is a failure. The AI cannot distinguish between different types of loss, projecting traumatic and unwanted memories (like her abusive father) along with the desired ones. This raises the central ethical problem: the device could be a "perseveration machine" that prevents healing rather than aiding it.
Step 04. Software Refinement via Valence Gating: The problem is identified as poor software modeling of grief. The team retrains the AI by having April explicitly tag unwanted memories, creating "negative spaces" in the data. This "valence-gating" approach works, teaching the system to filter for summon-worthy grief and ignore other negative affective states.
Step 05. Stabilizing Memories with Auxiliary Input: The projections are now correct but unstable, flickering out after a few seconds. The team concludes a single person's memory is insufficient. They conduct group sessions, gathering memories from April's friends and family to create a more robust, cross-validated dataset. This "auxiliary memory" training successfully stabilizes the projections for several minutes.
Step 06. Patient Ground-Truthing & Psychological Risk: April undergoes final implantation. The technology works as designed. However, a post-op check-in reveals she has withdrawn from life, consumed by the visions. After an intervention, she eventually recalibrates and learns to use the implant as a healthy tool for mourning. Despite this positive outcome, the narrator is convinced the psychological risk is unacceptably high and decides the project was a mistake. He moves on to less fraught applications for neuro-technology. I don't care to speculate on what others might do with this tech, but the narrator's conclusion seems sound.