The Story So Far

What a House Finch nesting cycle actually looks like, in data — as it unfolds. This page covers Day 1 (2026-05-29) → Day 36 (2026-07-03), from a pair returning to the wall sconce through fledging. Every chart below regenerates from the same daily-metrics JSON the journal entries are built from. As of Day 36 (Jul 3), the four nestlings are ~16 days post-hatch and in the 11–19-day fledge window (mode 12–15) — fully feathered and branching out of the cup. The first full daylight with an empty cup is fledge.

20 fully-observed days + 12 partial across a 36-day span (4 dark days, no footage) · 14,948 motion clips · 179 two-bird frames · current phase: Fledging.

How to read these charts. Not every day was filmed end-to-end, and one stretch wasn't filmed at all — so a dip can be the camera, not the birds. Three kinds of day appear below:

Full — the camera covered ~the whole day (≥20h of span). Solid bars, filled points.
Partial — the camera filmed only part of the day (setup, a hatch day that stopped mid-afternoon, the outage's first and last hours). Drawn translucent / hollow so a short bar isn't misread as a quiet nest. These are: Day 1 (May 29, 0h), Day 2 (May 30, 9h), Day 3 (May 31, 13h), Day 4 (Jun 1, 17h), Day 12 (Jun 9, 12h), Day 13 (Jun 10, 8h), Day 20 (Jun 17, 15h), Day 21 (Jun 18, 15h), Day 29 (Jun 26, 3h), Day 34 (Jul 1, 1h), Day 35 (Jul 2, 15h), Day 36 (Jul 3, 15h).
Gap — no footage exists for that day (the Jun 27–30 network outage). Shown as an empty labelled slot, never spliced out, so the hole in the record stays visible. These are: Day 30, Day 31, Day 32, Day 33 (Jun 27–30).

Coverage is measured by how much of the day the camera actually filmed (the span from first to last clip), so it catches days that started late or stopped early; a rare outage in the middle of an otherwise full day would not be flagged. Headline totals count full and partial days separately and never invent data for the gap. See The Gap for the outage write-up.

Phase timeline

Each bar is one day, sized by motion-clip volume and colored by the biological phase the nest is in. As the cycle moves from building → laying → incubation → hatching → brooding → fledging, the colors shift; the heights tell you how busy the cup was on each day. Translucent bars are partial-coverage days; the blank slots (Days 30–33) are the outage.

Motion volume

How many motion clips the Tapo captured each day, split by what the classifier saw. Bird-in-frame and no-bird stack over an “unclassified” band for the early days (before the classifier came online or while the GCP project was in billing dunning), so every bar’s full height is real motion. Translucent bars are partial-coverage days; blank slots are the outage.

Cup contents

Two things at once. The faint lines are the classifier’s floor — the most eggs and the most chicks the camera could clearly count in any single open-cup frame that day. A brooding adult hides the cup most of the time, so those undercount badly: the chick floor never gets past 3 even though the brood was five. The bold confirmed brood line is the eye-verified truth from data/brood.toml — five nestlings from the Jun 17 hatch, dropping to four when one was ejected on Jun 22, and holding at four since. That four is the number fledging watches: the day the cup goes empty in daylight, all four have gone.

What the cup held

Every classified clip sorted into one state of the cup: an adult brooding with the contents hidden, an adult beside visible eggs or chicks, eggs or chicks alone with no adult, or a truly empty cup. The stack shows the daily mix — adult-on-cup dominates incubation; the chick bands grow in after the hatch.

Who’s on the nest

Clips per day sorted by which adult the classifier saw at the cup. House Finch incubation is female-only, and it shows — the female line towers over everything from Day 11 on. The male appears mostly at the rim (see courtship feeding), and the “both adults” line is the tell: it spikes around Day 19 as the pair synchronise for the hatch, then holds up through provisioning when both feed the brood.

Courtship-feeding pulse

House Finch males don’t incubate — they provision the female at the rim of the cup with food. This line counts the frames each day where both adults appear together: the male’s arrival, the food transfer, his departure. It climbs across incubation, spikes just before the hatch (Day 19), then settles as the male shifts from feeding the female to feeding the chicks. The line breaks across the outage rather than connecting Day 29 to Day 34.

Nest care & provisioning

Feeding is the behavior that separates incubation from brooding. Three signals here: clips with food visible in the beak, clips where an adult perches on the cup edge (the feeding posture), and the classifier’s feeding-visit flag. All three sit near zero while the female is incubating eggs, jump at the Day 19–20 hatch, and stay elevated as both parents provision — the clearest data signature of the brooding→provisioning crossover.

Quiet cup

Frames where the camera fired but no bird was in view — the cup as the camera sees it without an adult on it. During building these are short windows between material-carrying trips. During laying they get rarer as the female commits more. During incubation they collapse to nothing. Once the chicks fledge, the cup goes quiet for good — so this is the chart to watch for the empty-cup day that marks fledging.

Weather

What the camera was filming through. Lafayette, California — daily high and low temperature with the diurnal swing shaded between, max wind on the right axis, precipitation marked as bars when it falls. Useful when reading the visit-frequency charts back: hot afternoons drop activity, rain compresses it into the dry windows.

Phase details

Where the cycle has been, where it is, where it’s going. The status column updates from data/stages.toml — the moment Casey edits that file (or the classifier flips a stage), the page re-renders.

Phase	Status	Start	End	What it is
Nest-building	complete	2026-05-29	2026-06-01	Female pressing material into the rim of the wall sconce. Male tracking, courtship-feeding at the cup. Cup read as finished on Day 4 (2026-06-01).
Laying	complete	2026-06-02	2026-06-06	First egg confirmed in cup at 09:06 PT 2026-06-02 (Day 5). One egg per morning; clutch complete at 5 eggs on 2026-06-06 (Day 9).
Incubation	complete	2026-06-07	2026-06-17	Female-only. Began Day 10 (2026-06-07) on the complete 5-egg clutch — delayed onset for a synchronous hatch. Ran 11 days; the female brooded tight through the final night (Jun 16→17) and the first chick was out by dawn on the 17th — a few days ahead of the marked June 19–21 window.
Hatching	complete	2026-06-17	2026-06-18	Began before dawn on Day 20 (2026-06-17). The full five-egg clutch was intact at 01:18 PT; the first hatchling was out by 04:45 PT (four eggs + one chick, confirmed from the top-down camera). The rest of the clutch followed through the day; by Day 21 all five had hatched — a brood of five, confirmed by eye. Synchronous hatch on a 5-egg clutch, a few days ahead of the marked June 19–21 window. (Note: the classifier's wide-frame egg counts undercounted the occluded cup.)
Brooding → Provisioning	complete	2026-06-19	2026-07-01	Five eggs hatched Jun 17–18 (Days 20–21). One nestling was lost on Jun 22 — worked out of the cup and ejected from the nest — leaving four; cause undetermined. The four nestlings passed the brooding→provisioning crossover (days 7–9 post-hatch) in the last week of June: the female now broods only intermittently and both adults provision at pace. By Jul 1 (~14 days post-hatch) the chicks are feathered and near-fledging — inside the 11–19-day fledge window. A five-day capture outage (Jun 27–30) fell across this stretch, so the late-brooding data has a gap; footage resumed Jul 1.
Fledging	active (now)	2026-07-02	—	Underway. As of Jul 2 (~15 days post-hatch) all four nestlings are fully feathered and have branched out of the cup onto the sconce ledge — the last stage before they leave. House Finch young fledge 11–19 days post-hatch (mode 12–15), so this brood's window runs Jun 28 → Jul 6. Exact branching onset is uncertain: the Jun 27–30 capture outage sits inside the window and footage only resumed Jul 1. Fledge itself is the first full daylight with an empty cup — watch the live stream and the Quiet-cup chart; the finale could happen on camera any day now.

Raw data

Each chart on this page is built from a small JSON file per day. The journal entries are built from the same files. They’re plain JSON; the camera saw what it saw. Days marked partial filmed less than 20 hours; the four dark days have no file at all.

Day 1 (May 29): 2026-05-29.json partial · 0h
Day 2 (May 30): 2026-05-30.json partial · 9h
Day 3 (May 31): 2026-05-31.json partial · 13h
Day 4 (Jun 1): 2026-06-01.json partial · 17h
Day 5 (Jun 2): 2026-06-02.json
Day 6 (Jun 3): 2026-06-03.json
Day 7 (Jun 4): 2026-06-04.json
Day 8 (Jun 5): 2026-06-05.json
Day 9 (Jun 6): 2026-06-06.json
Day 10 (Jun 7): 2026-06-07.json
Day 11 (Jun 8): 2026-06-08.json
Day 12 (Jun 9): 2026-06-09.json partial · 12h
Day 13 (Jun 10): 2026-06-10.json partial · 8h
Day 14 (Jun 11): 2026-06-11.json
Day 15 (Jun 12): 2026-06-12.json
Day 16 (Jun 13): 2026-06-13.json
Day 17 (Jun 14): 2026-06-14.json
Day 18 (Jun 15): 2026-06-15.json
Day 19 (Jun 16): 2026-06-16.json
Day 20 (Jun 17): 2026-06-17.json partial · 15h
Day 21 (Jun 18): 2026-06-18.json partial · 15h
Day 22 (Jun 19): 2026-06-19.json
Day 23 (Jun 20): 2026-06-20.json
Day 24 (Jun 21): 2026-06-21.json
Day 25 (Jun 22): 2026-06-22.json
Day 26 (Jun 23): 2026-06-23.json
Day 27 (Jun 24): 2026-06-24.json
Day 28 (Jun 25): 2026-06-25.json
Day 29 (Jun 26): 2026-06-26.json partial · 3h
Day 30: no capture (Jun 27–30 outage)
Day 31: no capture (Jun 27–30 outage)
Day 32: no capture (Jun 27–30 outage)
Day 33: no capture (Jun 27–30 outage)
Day 34 (Jul 1): 2026-07-01.json partial · 1h
Day 35 (Jul 2): 2026-07-02.json partial · 15h
Day 36 (Jul 3): 2026-07-03.json partial · 15h

On the counts: the classifier's egg and chick counts are conservative — it only counts what it can clearly see in an open cup, so they read as a floor, not a census (a brooding adult hides the contents most of the time; across the whole brood period the classifier never counted more than three chicks even though there were five). That is why the Cup-contents chart overlays the eye-confirmed brood (five from the hatch, four after the Jun 22 ejection) from data/brood.toml — the authoritative count, edited by hand when it changes, that fledging detection relies on. The per-frame chick_frames signal is gated to zero before the confirmed hatch, because every pre-hatch "chick" is the incubating female misread in a dim, mostly-occluded cup; after the hatch it reflects clips with a visible nestling. On coverage: bars and points are drawn solid for fully-observed days, translucent/hollow for partial-coverage days (<20h filmed), and left blank where no footage exists (the Jun 27–30 outage) — so a short bar never gets mistaken for a quiet nest. The motion and two-bird charts use confirmed counts throughout.

Auto-generated from static/data/daily-metrics/*.json + data/stages.toml at 2026-07-03T22:08:34.288Z by scripts/generate-story-so-far.ts. Re-renders on every daily publish (15:00 PT) and on stages.toml change.