FIELD RESEARCH · WORKING NOTES, NOT MARKETING CLAIMS
KNOW
THE PANEL.
Research that directly changes the architecture. Facts are linked to their sources; project consequences are clearly marked as our interpretation.
R01
A pixel transition is a sequence
An electrophoretic pixel is not assigned a color once. A controller applies a temperature-sensitive sequence of voltage states based on the old and desired state. That lookup sequence is the waveform—and it is why a compositor must remember what the panel believes is on screen.
Design consequence: updates carry source state, target state, region, mode, and temperature context.
SOURCE · E Ink 9.7-inch kit manual ↗R02
Fast and clean are different jobs
A direct update can change affected pixels to black or white without the full clearing flash. A grayscale clear can deliver the best image appearance while touching the whole display. Neither should be the universal default.
Design consequence: widgets express fidelity and urgency; policy chooses the supported waveform.
SOURCE · E Ink display mode guide ↗R03
Dither is content policy
Binary interfaces can remain crisp and fast while photographs trade spatial detail for perceived tone. Dither should be opt-in by surface and content class, not smeared across text, rules, or icons by a general-purpose renderer.
Design consequence: the render tree labels photographic regions; text remains native 1-bit.
SOURCE · Modos Glider technical notes ↗R04
The controller changes the ceiling
Some controllers support runtime-selected regional modes, internal state, and accelerated dithering; cheaper modules may expose only constrained SPI commands. Device agnosticism must mean honest capability negotiation, not lowest-common-denominator fiction.
Design consequence: profiles advertise region alignment, queue depth, modes, color states, and cleanup rules.
SOURCE · Modos Glider controller research ↗R05
Color is not one technology
Color-filter-array panels and multi-pigment systems make very different compromises. Recent Spectra waveform work aims to reduce flashing and expand mixtures, but black-and-white interaction remains the predictable baseline for the OS.
Design consequence: color is a panel capability and an explicit refresh class, never a hidden CSS assumption.
SOURCE · E Ink Ripple announcement ↗R06
Linux is a foundation, not the UX
Open devices such as PineNote show both the leverage of a Linux userspace and the difficulty of system-wide refresh controls, waveform switching, dithering, suspend, and proprietary display paths. The project can reuse Linux while replacing desktop assumptions.
Design consequence: isolate panel control behind one auditable service and design suspend from day one.
SOURCE · PINE64 development notes ↗R07
AI may draft a driver. It cannot certify one
Model-generated kernel and firmware code can accelerate register-map transcription, fixtures, and first-pass drivers. It is still untrusted code at the most privileged layer: hallucinated timing, polarity, voltage, or suspend behavior can corrupt data or damage hardware.
Design consequence: generated code must stay behind review, KUnit or kselftest coverage, emulation where possible, current-limited bring-up, and named-device hardware traces before it enters a supported profile.
SOURCE · Linux kernel testing guide ↗