Dichoptic Rendering on Consumer Hardware: What AB 1709's Screen-Time Fight Misses About Attention

Google, Meta and TikTok sued California this week to block AB 1709 — a law that bans platforms from giving under-16s "psychologically exploitative features" like autoplay and algorithmic recommendations that drive compulsive use. The platforms argue the law overreaches. The state argues the features are engineered to keep kids hooked.

Notice what nobody in that fight disputes: that the visual and attentional systems of a young person are plastic, and that what you point a screen at matters.

That second half is the interesting part for anyone who builds on consumer hardware. Because the same headset, the same video feed, and the same attention-shaping techniques being litigated can be turned 180 degrees and pointed at a therapeutic problem instead of an engagement metric.

One place that's already happening is amblyopia — and it turns out the engineering is genuinely interesting.

The problem the old default leaves on the table

"Lazy eye" is a mismatch. One eye sends a stronger signal than the other, the brain learns to prefer it, and depth perception quietly degrades. The classic treatment — occlusion with an adhesive patch — works, but it's passive, uncomfortable, and heavily front-loaded onto childhood. Adults are often told the window has closed.

The mechanism underneath the patch, though, isn't about age. It's about giving the weaker eye a slightly harder job and rewarding the brain for using both eyes together. That's a signal-processing problem as much as a clinical one.

The rendering trick: two feeds, one scene

A VR headset is, from an architecture standpoint, the ideal substrate for this. Two independent eyepieces mean you can render a different image to each eye from a single decoded video frame. That's dichoptic presentation, and on a Meta Quest it's essentially a per-eye post-processing pipeline rather than a hardware problem.

Amblyotube, developed by Seven Sports, is recreational and educational software for visual coordination and attention exercises in VR. It runs on Meta Quest and shows a different visual experience to each eye while a YouTube-style video plays. The interesting part is what each stage of that pipeline actually does:

• Dominant Eye Shader. The stronger eye gets a gentle, adjustable digital blur — plus contrast, brightness, gamma, and opacity controls. Not a blackout. The point is to shift the relative information value of the two eyes without breaking fusion entirely. At 100% opacity it becomes full digital patching, but the default posture is partial transparency.
• AI Sharpener (MFBF mode). On the lazy eye's side, an engine detects human figures in the video and applies selective sharpening plus visual accents — bold red outline, yellow-green highlight, pulsing at an adjustable frequency. It gives the weaker eye a high-salience target to lock onto. Controlled flicker is the same idea applied temporally.
• Wedge Occlusion. Part of each eye's view is blacked out so each eye sees unique portions of the scene. This attacks binocular rivalry at the source — the tug-of-war that makes fusion hard in the first place.
• Magenta Focus Cue. A bright magenta cue in the lazy eye's view plus a subtle grey dot in the dominant eye's view. Magenta is a deliberate choice: it's rare in natural video (skin tones, greens, blues), so it stays high-contrast against almost any frame. As the two cues track together, the brain is pushed to merge them into one cohesive 3D image. That's the fusion step.

No adhesive patches. No red-blue glasses, which matters because anaglyph approaches distort color across the whole scene; this keeps natural color intact.

Configuration is the part that bites

Here's where the engineering meets the clinic. On launch you select which eye is affected — Left or Right. Getting this wrong trains the dominant eye, so it's a hard fail rather than a soft one. From there you pick a condition preset while wearing corrective glasses if applicable: Mild (can't read the last 2 lines of an eye chart), Moderate (last 3–5 lines), Severe (5 or more). The preset auto-tunes the default levels for the Dominant Eye Shader and Lazy Eye Sharpener.

There's also an integrated LogMAR vision check: Landolt C optotypes at a virtual 3-meter distance, up to 30 trials per eye, using a clinical staircase algorithm. It reports logMAR acuity scores per eye and for both together, and can automatically update the therapy profile to match the detected severity. That's a nice closed loop — measure, set parameters, train — that usually requires a separate appointment.

What the lawsuit frame gets wrong

AB 1709 is a fight about whether attention-shaping features should be pointed at minors without limits. Fair fight. But it presumes the only possible outcome of a screen session is extraction.

Amblyotube is the counterexample in miniature: the same 30–40 minutes (max 1 hour) a day that would otherwise be pure consumption becomes an active vision gym, because the software is optimizing for binocular coordination instead of watch time. Same hardware. Same content. Different objective function.

Practical notes for anyone evaluating it: ages 13 and up, recommended 30–40 minutes daily with a 1-hour maximum per session to avoid VR eye strain. It supports Meta Quest controllers and hand tracking, has a floating resizable control panel, integrated YouTube search, and an offline catalog of child-friendly videos — useful if you want a session without a network dependency. Room View (AR pass-through) streams over your real room as an alternative to the default virtual environment.

The honest limits

To be clear about what this is: Amblyotube is a supplementary vision-training and assistive tool for binocular coordination and depth perception. It is not a diagnosis, a treatment, or a cure. It's designed for people who want to improve visual coordination and attention, including those with lazy eye or amblyopia. If that's you, talk to an eye-care professional about whether this kind of training fits your situation.

If it does, it's a download away: https://www.meta.com/en-gb/experiences/amblyotube/25906906972338493/