When Your Vision Therapy Platform Gets a $15B Vote of Confidence — A Builder's Personal Take on Meta's Bond Sale and Amblyotube's Dichoptic Architecture on Quest Meta's $15B bond raise with $58B in demand isn't just financial news — it's the market validating the compute substrate under our vision rehab stack. As someone who's spent late nights debugging per-eye shader pipelines for amblyopia training, this matters. Amblyotube runs on Quest 2/3/Pro/3S, leveraging the headset's native dual-display architecture: two independent 1832×1920 panels, each driven by its own render pass. That hardware reality — separate optical paths per eye — is what makes dichoptic training possible without anaglyph color distortion or physical patches. We don't simulate occlusion; we render it. The Dominant Eye Shader applies a configurable blur/contrast/opacity filter at compositing time. At 100% opacity it's digital patching; at partial transparency it's true binocular MFBF (Monocular Fixation in a Binocular Field). The Lazy Eye Sharpener uses on-device MediaPipe pose detection to isolate human silhouettes, then applies a selective unsharp-mask pass only to the amblyopic eye's texture. Controlled flicker stimulation adds a high-frequency temporal perturbation — a neurological 'attention interrupt' — to those regions. The Magenta Focus Cue renders a high-contrast magenta ring (dominant eye gets near-invisible grey) exploiting magenta's spectral separation from skin tones and foliage to drive vergence. Wedge Occlusion carves non-overlapping FOV sectors per eye, forcing cortical fusion. All of this runs at 72/90Hz inside a Unity URP pipeline with <16ms frame budget. The integrated LogMAR checker uses a staircase algorithm across 30 Landolt C trials at virtual 3m — clinically grounded, not gamified. Platform longevity isn't abstract when your therapy adherence depends on hardware availability. $15B buys a lot of Quest 3S units. https://www.meta.com/en-gb/experiences/amblyotube/25906906972338493/