The Screenless Sight Codex

Hypothesis Registry

This registry turns broad ideas into researchable or modelable hypotheses. It does not ask whether an idea is “allowed” to exist. It asks how to sharpen it.

Maturity Scale

Level Name Meaning
H0 Metaphor Useful analogy, not yet a mechanism.
H1 Mechanism Sketch Names a target, carrier, and transducer.
H2 Toy Model Can be represented in diagrams, simulations, or human-interface demos.
H3 Prototype Analogue Can be explored with ordinary screens, audio, haptics, or nonclinical devices.
H4 Research Parallel Resembles active scientific or clinical research.
H5 Translation Candidate Could become a practical system if major engineering and evidence gaps close.

Core Hypotheses

ID Hypothesis Level How To Make It Sharper
H-001 A scanned stimulus can produce useful perception without a conventional screen. H4 Compare raster scanning, retinal scanning displays, and cortical shape tracing.
H-002 Artificial vision may work better as a learned language than as bad natural vision. H3 Build symbol sets, train users, and compare recognition speed.
H-003 Sparse semantic cues can outperform low-resolution images for navigation tasks. H3 Test scene skeletons against blurred video and audio descriptions.
H-004 A personal percept map is the real display surface in neural prosthetics. H4 Track perceived location, shape, and stability over time.
H-005 Event-based sensing is a natural fit for low-bandwidth visual substitution. H3 Compare frame-based versus motion-change cue streams.
H-006 Foveated artificial vision can reduce bandwidth without losing task usefulness. H3 Use gaze or attention models to allocate detail.
H-007 A phosphene font can make prosthetic percepts more useful than pixel imitation. H2 Define a symbol vocabulary and train recognition.
H-008 A retinal interpreter should mimic useful retinal transformations, not raw pixels. H2 Encode contrast, edges, motion onset, and adaptation.
H-009 Cortical stimulation may benefit from temporal tracing rather than static patterns. H4 Compare static symbols with traced shapes in literature and simulations.
H-010 Cross-modal redundancy can make weak visual channels more reliable. H3 Pair visual-like cues with audio or tactile confirmation.
H-011 Remote carriers become more plausible when they communicate with local transducers. H1 Specify the local receiver, conversion mechanism, and addressing model.
H-012 Memory-adjacent systems should begin as voluntary cueing systems. H3 Compare images, narration, location, and repetition as recall supports.
H-013 A “neural shadow mask” can be modeled as crosstalk control. H2 Simulate spread, overlap, and separation in a percept grid.
H-014 Biological-phosphor concepts become sharper when classified by input/output conversion. H1 Compare phosphors, optogenetic proteins, fluorescence, and implantable converters.
H-015 A semantic subtitle track may be the most practical near-term screenless layer. H3 Rank cue types by usefulness, interruption cost, and error risk.
H-016 The trained perceptual loop can function as the effective medium of a screenless interface. H3 Compare passive cue viewing with active exploration, training, and task performance.
H-017 Removing middle representational layers can improve usefulness when the remaining cue grammar is stable. H2 Compare full low-resolution imagery against edge, motion, depth, label, and scene-graph cue sets.
H-018 Transparent assistive tools should be judged by action fluency, not image fidelity alone. H3 Measure navigation, reaching, recognition, and user confidence after practice.

Hypothesis Expansion Template

## H-000: Hypothesis Name

- Maturity:
- Related scenarios:
- Closest known science:
- Target:
- Carrier:
- Transducer:
- Encoding:
- Toy model:
- What would count as progress:
- What would falsify or weaken it:
- Open questions:

Toy Model Ideas

Toy models are useful because they let the project explore bold ideas without waiting for future hardware.

Toy Model What It Tests
Low-resolution phosphene simulator Whether sparse light dots can support recognition.
Shape-tracing simulator Whether temporal drawing beats static dot patterns.
Scene skeleton annotator Whether sparse spatial graphs answer practical navigation questions.
Phosphene font trainer Whether a small artificial alphabet becomes fluent.
Event-stream visualizer Whether changes and motion carry enough information.
Cross-modal cue trainer Whether audio and tactile cues improve weak visual symbols.
Memory lantern card Which voluntary cues support later recall.
Retinotopic bus diagram How location, feature, timing, and confidence might be separated.
Neural shadow-mask simulator How crosstalk control changes symbol clarity.
Perceptual codec bench Whether fewer task-first cues can beat noisy full-image displays.
Active scan trainer Whether user control makes sparse cues become perception-like faster.
Confidence grammar test Whether visible uncertainty improves trust and decision-making.

High-Value Research Questions