Per keystroke, diff the new frame against the last shown one to find the
changed row band and partial-refresh only those rows instead of all 272.
E-paper update time scales with the gate rows driven, so a single edited
text line refreshes far faster. Windowed in Y only (full width, both
controllers), so the seam/mirroring logic is unchanged. Also drop the
periodic full refresh to every 64 updates — the panel stays visually
clean, so it's now mainly for longevity.
Spike 5 — first spike where keyboard input and panel output run
together. usb_kbd changes from a blocking run() into start(): it brings
the USB host stack up on background threads and pushes edge-detected,
US-layout-translated key-downs onto a queue drained via next_key().
main.rs owns the panel, maintains a wrapped/scrolling text buffer, and
partial-refreshes per keystroke batch with a periodic full refresh to
clear ghosting. Logs per-refresh latency.
display_frame_partial writes the new image to bank 0x24, runs GxEPD2's
partial waveform (_Update_Part: 0x3C/0x80, 0x21/0x00,0x10, 0x22/0xFF,
0x20), then syncs bank 0x26 so the next partial has a correct baseline.
Like GxEPD2 for this dual-controller panel, the update covers the full
panel with the partial waveform — no windowing, since the waveform time
dominates. Much faster than a full refresh and without the flashing.
Spike 4. Drive the ESP-IDF USB Host Library directly through the raw
esp-idf-sys bindings (the convenience HID class driver is a managed
component not vendored in mainline, and a boot keyboard doesn't need
it). On attach, src/usb_kbd.rs enumerates the device and dumps its
descriptors, claims the boot-keyboard interface, sends SET_PROTOCOL(boot)
and SET_IDLE(0), then polls the interrupt-IN endpoint and decodes each
8-byte report into modifiers + keycodes logged over UART.
main.rs becomes the Spike 4 harness; the epd module is kept compiled
(allow(dead_code)) so it doesn't bit-rot. Verified on hardware with a
19f5:3255 keyboard: letters, digits, modifiers, and rollover all decode.
Add a thin dual-SSD1683 driver (src/epd.rs, ported from GxEPD2's
GxEPD2_579_GDEY0579T93) for the 792x272 e-paper panel: reset/init,
RAM addressing, the split-and-mirror full-frame blit across the
master/slave seam at x=396, and full-refresh waveform. Frame
implements embedded-graphics' DrawTarget.
Replace the Spike 1 blink harness in main.rs with the Spike 2 test:
SPI at 4 MHz over SCK 12 / DIN 11 / CS 7 / DC 6 / RST 5 / BUSY 4,
alternating normal/inverted frames with a seam-straddling circle,
"Typoena", and the build tag. Verified on hardware 2026-07-04.
Emit BUILD_TIME and BUILD_GIT (git describe --always --dirty) as
rustc env vars so the running firmware can identify itself on serial
and on-panel. A rerun-if-changed on a nonexistent file forces the
script every build, keeping the timestamp fresh. Bring-up lesson:
know which build you're diagnosing.
The bench board is a DevKitC-1 v1.0 (RGB LED on GPIO 48), not the v1.1
layout (GPIO 38). Vendor the official v1.0 pinout diagram under docs/ and
reference it from the README so the pinout is pinned offline, and fold in
the 2026-07-04 on-hardware blink verification note.
The DevKitC-1 has no discrete LED on GPIO 2, so the Spike 1 blink was
invisible without external wiring. Drive the on-board addressable LED
(WS2812 on GPIO 48) each cycle for visible confirmation on the bench.
Zed's bundled rust-analyzer calls `cargo metadata --lockfile-path`,
which is still gated behind `-Z unstable-options` on cargo 1.95.0
(both stable and the esp Xtensa fork). Workspaces fail to load with
the bundled binary.
Point lsp.rust-analyzer.binary.path at the rustup-managed
rust-analyzer (version-locked to cargo 1.95 stable, no unstable
flags). Also set cargo.target = xtensa-esp32s3-espidf with
allTargets=false so RA doesn't try to check the crate for the host
target — esp-idf-sys can't build for macOS.
Generated from esp-rs/esp-idf-template for the ESP32-S3 std target.
src/main.rs toggles GPIO 2 every 500 ms and logs `blink N` over USB-
serial — the minimum bring-up surface called out in
docs/v0.1-mvp-technical.md (Spike 1: confirm toolchain, flash, and basic
GPIO). edition=2024 with rust-version=1.85.
No editor/render/git/usb/fs modules yet; those land per the spike
methodology when later spikes need them.
Latency is the dominant signal for W1 (sub-second response to typing)
and on a typing-focused device the reMarkable 2 + Type Folio's e-ink
delay disqualifies a 2. Updates the TikZ perception array, the table
row + rationale, and the totals narrative (reMarkable 45 → 44, ranking
unchanged).
Two stale bits in §4's preamble. (a) "15×15 roof matrix" still
referenced H14's pre-retirement column count — corrected to 14×14.
(b) The symbol legend named classical QFD glyphs (◎ ○ × ⊗) but
quality-house.md's roof actually renders ASCII (++ + − −−). Both
listings now lead with the ASCII glyphs that match the source of
truth, with classical equivalents kept in parens.
GLOSSARY.md was reachable only via docs/qfd.md's §2 preamble.
Surfacing it from README adds two entry points: the intro and
software-stack pointers to docs/qfd.md now follow up with a link
to GLOSSARY.md (the ontology layers used across the docs), and
the repo-layout block lists GLOSSARY.md alongside CONTEXT.md so
the two glossaries' division of labour — device vocab vs
methodology vocab — is visible at a glance.
The H10-H15 mild conflict node still pointed at C-10-15 (H15's
pre-retirement column). H14's removal renumbered H15 to column 14,
so the roof coordinate becomes C-10-14. Missed in the earlier
sweep — comment line already says "H10-H15" so the intent was
unchanged, only the cell coord was stale.
The doc previously conflated functions (transformations the device
performs) with characteristics (measurable attributes). Surfacing
Functions as their own ontology layer makes HOW names — which pack
a function reference plus an attribute — readable without function
definitions staying implicit.
- new GLOSSARY.md at root: 5-layer ontology stack (WHAT / Function /
Characteristic / Metric+Unit / Target), peer to CONTEXT.md
(device-specific vocabulary)
- qfd.md §2 opens with a Functions inventory (Type, Save, Publish,
Recover, Boot, Provision); Render and Reconnect noted as
sub-functions; Provision flagged as build-time-only in v0.1
- H1 "Keypress → glyph latency" → "Type latency (keypress → glyph)"
- H4 "Cold boot → cursor ready" → "Boot latency (cold)"
(the arrow transformation moves to the Functions inventory; the
parenthetical keeps the residual qualifier)
- cascade through §3 priority list, §6 critical-performance-budget
rows for H1/H4, §8 inconsistencies, and quality-house.md TikZ
column labels
Matrix cell strengths held; no Σ recompute.
Module count / public-API surface is a static source-code property,
not a runtime function nor an artifact characteristic like H10 binary
or H15 build time. It was a refactor-leverage proxy for W9 — proxies
of proxies belong in the ADRs and the component matrix, not in the
HoQ. Removed from §2, §5 matrix, §4 conflict list, and the C12
overloaded-list mention; basement Σ total drops 1674 → 1557 so rel%
recomputed in quality-house.md.
The post's job is to land the timing question and the apocalypse
reframe — Wi-Fi rationale, optimisation diversions, and generalising
reflections were overexpansion. Cut.
The user-meaningful moment of a Publish is when the local commit
lands (~0.2 s), not when the push completes (~5–10 s). New principle
in CONTEXT.md, instantiated in v0.1 product doc with staged status
transitions and a corrected on-demand Wi-Fi status example.
Long-form rationale for the "durability before delivery" framing —
why a 10-second network operation in response to a single keystroke
doesn't have to feel slow. Establishes docs/notes/ as the home for
shareable design rationale (not authoritative spec).
Always-on station mode burns ~410 mAh/day on the radio alone; on-demand
drops that to ~25 mAh/day at 10 Publishes/day. Radio comes up on Ctrl-G
with a 10 s association timeout, holds for a ~90 s grace window after
each Publish, then tears down. SNTP cached up to 24 h to avoid
re-querying every Publish.
Also removes stale SoftAP provisioning bullets that contradicted the
"v0.1 has no provisioning module" stance later in the same doc.
The user's mental model of Publish is a Google Doc backed by git, not a
curated commit log. Branches fall out of scope under that framing; the
device tracks one stream of work on whichever branch the remote was
cloned on, and never switches. Drops :Gbranch from v0.7 to match.
Captures "I can carry the device and write away from a desk" as a
distinct WHAT from W11 (multi-day battery), weight 8. Recomputed
basement Σ (1546→1674); H8 climbs #6→#3 in the §3 priority list as
its voter base widens to include W14, and H12 enters the top six at
#4. §6's "(b) narrow voter base" override for H8 is retired and a new
§7 ADR-008 vs W11+W14 unresolved-conflict bullet records that wall
power in v0.1 now disappoints two WHATs, not one.
reMarkable 2 with Type Folio tested less responsive than Freewrite Smart
Typewriter (Sailfish). With Smart at 3, reMarkable drops one notch to 2.
Total: reMarkable 42 → 41, now tied with Smart at 41.
Sailfish ships on Smart Typewriter Gen 3 as well as Traveler, so the
same firmware-driven bumps apply: W1 2→3 (Smart still trails Traveler
by one notch on panel size), W5 3→4, W9 1→2. Smart total 38 → 41.
Project gets a proper name. Replaces "Ours" / "Our product" placeholders
in the QFD perception zone and the working title "typewriter" in
README/CONTEXT headers and the npm name. "Typewriter" the category
noun (the writing-machine concept) stays unchanged — Typoena is the
brand, typewriter is the category. SoftAP SSID typewriter-setup and the
repo directory are deliberately left as-is.
The Sailfish firmware (2025-11-19) rewrote the Freewrite Traveler OS in
Rust, cut keystroke latency 40–100 %, and trimmed power draw -30 %
typing / -50 % idle. Three rows are bumped: W1 3→4, W5 3→4, W9 1→2.
Freewrite now ties reMarkable at 42; Pomera still #2 at 45.
text width={\qfdWhatW - 0.2}cm doesn't parse — the braces don't
evaluate the arithmetic, so TikZ sees "4.6 - 0.2cm" which isn't a
valid dimension. Pre-compute the narrower width with
\pgfmathsetmacro and reference the resulting macro instead.
Text box anchored at qfdLeftEdge + 0.1 with width qfdWhatW overflowed
0.1 cm past the cell's right edge into the Weight column, clipping
trailing words like "the", "lag,", "hob-". Shrink the box by 0.2 cm
so it has 0.1 cm clearance on both sides.
Spell out the Normal-vs-Insert distinction (relative with current line
absolute / absolute everywhere) so it's clear the gutter behaves like
Neovim's number+relativenumber combo, not flat absolute numbering.
Each product now gets a profile line across all WHATs using the same
colour and dash pattern its legend entry already advertises, so the
shape of each competitor's strengths/weaknesses is readable at a glance
instead of having to mentally connect the dots.
Raise the header zone (qfdHdrH 3.2 → 4.2) so rotated HOW labels stop
crossing into the roof diamond grid, and wrap the perception title onto
two lines via text width so it stays inside its zone instead of bleeding
left over the H14/H15 column labels.
Drop the duplicated 13x15 strength matrix and 15x15 roof tables from
qfd.md; the same data already lives in quality-house.md's TikZ source.
qfd.md §3/§4 keep the narrative reading (priority list, conflict list)
and point to quality-house.md for the cells and basement sums. Both
files' regenerate/maintenance notes flipped to reflect the new contract.
