Casey Tunturiarcane & algorithmic

74 repos into whatever /tmp the runner has is how the laptop run died at "no space left on device" — /tmp is RAM there. RUNNER_TEMP, /var/tmp if the runner does not set it.

Gitea clamps a page to MAX_RESPONSE_ITEMS — 50 by default — and ignores a larger limit without a word, so `limit=100` read as "all of them" and returned the first 50 of 74 repos. SouveraineOS, Pixel3Arch, souveraine-viewtop, souveraine-speech and souveraine-lens have been invisible to the ledger for its whole life; today alone it credited 2 repos where 7 had work in them, and the last ten days go 162 events to 413.

remove-section http does not touch URL-scoped subsection keys, so checkout's actions-user header survived and the push still resolved to the actions user. Unset the qualified key and the plain one.

TASK-36 said the load path had never executed. It ran 2026-08-11 — LoadNanoappResponse decoded, success byte set, and six gauges came off a flat baseline at 2502 and 2990 against a 2500 floor. Recorded with Casey's numbers and the two things that would otherwise compress away: the margin carries intent, and the proximity veto is a separate job from the threshold.

Record update delivery as a per-body offer and local candidate activation after a live shell push killed an active conversation.

Reconciles the two owning docs with souveraine cfad688. Sub-scope A was always half-landed: the substrate was multimodal end to end and the HTTP door was text-only, so the TUI could see and every other surface was blind. The door opens; nothing walks through it yet.

Firefox extension and Bitwarden popups were rendered above their parent but absent from client hit testing, so every contact fell through to the toplevel beneath. Walk application popups at the same origin and in the same parent-by-parent order used by rendering, including popup area outside the parent bounds.

the build reds whenever Google's CDN blinks — next/font fetched both families at compile time. now they live in the repo with their licenses; builds are hermetic. same css variables, same look.

the signaling build died on a DNS timeout resolving node:20-alpine, and the base was missing entirely though 1290 built with it minutes before — whatever ate it, a deploy must survive a cold cache. FROMs are digest-pinned now, and restart.sh pre-pulls the same digest with three chances. warm cache = zero registry contact; cold cache = the only time the resolver matters.

the old swap's 'down' found nothing because the live containers were born inside the old pipeline's job container — a different compose project namespace. rm by the names the compose file declares, then up. compose: drop obsolete version key

The composed config is symlinked into the repo and quickshell reloads on write, so an agent editing QML edits the live shell mid-turn. Stage changes as patches instead; the human applies and reloads on their own beat.

The Err branch surfaced the same 'no anomalies detected' text as the success path, at Low urgency. A subconscious erroring on every turn was indistinguishable from one noticing nothing. The only record of the real cause was a tracing::warn on a stderr whose socket has no reader.

The reload killer: FileView under a QtObject root — QtObject has no default property, so the whole service tree failed to compile. Every working service singleton roots in Quickshell's Singleton type.

The vault card lives where notifications were — header with sync pill (clean/ahead/behind/dirty), today's note entry, five freshest notes, cold start and push rows — over the notification list. Lens.qml owns souveraine-lens the way Face owns the rig: one socket, event lines up, no vault access of its own. Kiyoshi's slash grammar and tool registry are noted for the fuller app; the micro view only opens and plants.

The sidebar asks, the lens answers on the same line protocol: state returns the freshest notes and the git truth (ahead/behind now read from git status -sb); open drives the page by name; create plants a note and refreshes the page. Page gains thin __lens.open/refresh/sync/create wrappers so the shell never speaks page internals.

serve_ipc from the rig's discipline (stale-socket unlink, parent dir created); shell evals run in the page; page events leave as stdout lines the sidebar can hear. The vault stays the one thing with no other readers.

vault walker (frontmatter, wikilinks, graph), embedded page with a hand-rolled force layout, ipc bridge with no server connection, autocommit on save, house ci gates.

build_planes documented a wait it never performed, so with no exportable fence the atomic commit raced the renderer into the scan-out buffer. That is the tearing on the d10: simpledrm has no syncobj, supports_fencing is false, and nothing held the flip while llvmpipe was still writing.

build_planes already documented this: with no exportable fence, or a surface that cannot fence at all, the sync point has to be waited on. It never was, so the atomic commit raced the renderer into the scan-out buffer. Tears on simpledrm + llvmpipe, where supports_fencing is false and nothing held the flip.

Display pipe binds and the panel stays dark; touch worked once and stopped. Both have a named single-boot test rather than a theory. Driver changes and rails are ruled out by test; the display path is not.

blueline-screen-toggle does not exist here so the blank no-ops, and Dim is an absolute that reads as off on a 2047-step backlight. Lit-forever or black-forever, nothing between, until eb0292c ships.

One row per capability, one column per body, with a strict status vocabulary — "works" means exercised with evidence named, never that a device node appeared. blueline points at PAF/ rather than copying it; d10 carries its own state; m1 records what already transfers before anything is measured.

Touch proven with real events, not a device node: three layers, the last being runtime report activation gated to D11 so D10 went silent a second after boot.

The shell is on the iPhone 7 — viewtop, sessiond and qs, drawn by llvmpipe on a card with no render node, over one cable. Wallpaper, palette, bar, clock, dock and her quote match the Pixel.

STATE moved to 2026-08-09. One instrument is waiting on a finger; everything else about the enrol lifecycle is reachable in RAM today.

fprintd, gazed, sailfish-fpd and biometryd are the same interface four times. Take that shape and face and finger become one surface with two backends. Mapped every call onto the recovered TA commands; the gaps are persistence and the auth token, both already named.

The remaining citations in package recipes, overlay patches and task files still named the retired in-repo spine.

pkgrel 6. Adds a stream mode that holds both endpoint fds open and acknowledges each accepted report, so the shell-side controller is a persistent trackpad and keyboard rather than a process per keystroke.

SouveraineOS 72aaa53 copied the spine on 2026-07-24 and this repo kept a live copy, so both drifted through phone calls, Active Edge and the fingerprint work. The newer tree was promoted before this duplicate was retired; nothing was lost, and 9b1afd1 still holds the whole retired PAF.

fp-arm/bringup/go/load/sweep have been in ~ on the device since 2026-08-01; a reprovision would have taken them. fp-go's error table, decoded from the HAL, independently matches the router codes read out of the TA today -- recorded in the protocol doc.

Target 10 command 1 is check_finger_lost. Arms exactly once and brackets every query with the interrupt count, so the call's own edge is reported apart from spontaneous ones -- the 1444-event trap.

The only persistence is qsee_fts_*, reached from three wrappers. Call graph says targets 2 and 12 reach them and nothing else does, so the enrol/identify lifecycle runs in RAM with no listener. Parking TZ is the real hazard and it lives in target 2 alone.

The TA is not stripped: every handler logs __func__, so the router's jump tables resolve to names. Eight targets; sensor is 10 and bio is 11 with the full enrol/identify/enumerate/delete set. Target 10 command 1 is check_finger_lost, which is the instrument TASK-44 was missing. Static, from the vendor image on disk.

PersistentProperties is in-process, so it carries the surface choice through a reload and not through a fresh shell. sessiond locks before the shell exists, so a greetd-started shell adopts a live lock with Config.ready still false and binds the desktop surface — and WlSessionLock.surfaceComponent cannot change once active. A lock screen that will not take your PIN, on the iPhone 7 tonight.

brightnessctl set 10 is 4% on blueline's 255-step backlight and 0.5% on the iPhone 7's 2047-step Apple DWI, where it reads as fully off. Measured on d10 2026-08-09: brightness=1 with the panel still powered, so the screen went black on a dim that never blanked.

Power is the one irreversible transition and it bypassed the authority entirely: both menus called Session.poweroff() straight into systemctl, so there was no Action, no trail entry and no refusal path.

CI already built the x86_64 daemon and then threw it away — the binary and its user unit were copied into the package only under aarch64, so the laptop could never receive the thing every lock, blank and button report assumes is running.

78d6743 committed PowerMenu's half and never added the file, so the singleton in qmldir resolved to nothing and the shell would not load on a tree that had not been hand-linked.

Points at the personal agreement and the cross-repo map, names this tree as the substrate rather than a harness, and starts architecture at saf/INDEX.md.