Every frame before this is drawn on a build eight months old, and 8a's day one is really a build with a full budget and eight lines to place. A genuinely new project has neither side of the comparison this screen exists to make. Four states in the order they actually arrive: 9a nothing at all, 9b a budget and no spend, 9c the first few weeks, 9d where you are when fifteen of eighteen groups are untouched. 9e is day one at the desk. What holds them together: the screen never shows a verdict it cannot support, it never draws a run with nothing in it, and the reason it is empty is always a sentence naming what is missing and who has it.
This screen is a comparison, and on day one it has nothing on either side of it. No tabs, because two of the three would be empty and a tab bar promises content behind every one. No $0, because a zero measured against nothing is not a fact. One thing is waiting on the homeowner and it carries the only action; the block under it says what each reading needs before it can appear, in the order those readings arrive.
Now the zero is true, because there is something to measure it against, and a real total with an empty bar is more use than a hidden figure — it shows the shape of the build before it starts. The trend is waiting on time rather than on her, so it takes the dated container and no action: an action would imply she could hurry it. The projection stays at the outline rung and says why, which here is not a missing input but an arithmetic one.
Enough has happened to read two figures and not enough to read a verdict, and the title says exactly that rather than reaching for Steady. The support line shows the working — two per cent, three groups of eighteen — because a homeowner who can see why it is silent will trust it when it speaks. The tabs arrive here: two of the three now have content behind them.
The third tab, early. Eighteen rows of which fifteen say nothing would be a list the eye cannot hold and a screen that looks like a fault, so only the groups with something against them are drawn and the rest are one counted sentence. The counts are counts, never a fraction of the build. Each row shows the estimate it is measured against, because a figure with nothing beside it is the thing this screen replaces.
The desk form of 9a, and the width is a liability rather than an opportunity: three blocks stretched across 1160px read as a screen that failed to load. The pane keeps one column for the thing waiting on her and pairs the two explanations beside it, which is the only two-up on the screen because they are the same kind of thing. Compare with 8a, which is this screen one budget later.
Of the seventeen figures 7b draws, two can be served today. 8a is the screen with only those two on it — a designed not-at-all rather than a degraded 7b, on the same principle the timeline already follows: return no estimate rather than part of one. 8b is the fourth verdict state, where an unresolved disagreement between two documents suppresses the reading entirely. Both of them say what is being waited on and who can end the wait. 8c and 8d draw the close-out overlay open, phone and desk — it is live on 7a and 7b behind "Mark one of these finished", and a state only reachable by tapping is one nobody reviews. 7a and 7b have also gained the unmatched bucket, a row for spend whose cost code is null.
The overlay drawn open, because a state you can only reach by tapping is a state nobody reviews. It is the same sheet 7a opens from "Mark one of these finished" under the still-moving run. The manual door matters most to the homeowner it is least visible to: with no builder connected and no draws, nothing exists to infer a finish from, so the proposal in the weekly queue never fires and this is the only way settled ever gets set.
The desk form, drawn on its own rather than inside 7b's shell where it lives. Past the tablet rung a sheet comes from the side, and it is bounded rather than full height so the title, the four names and the action that commits the choice stay in one eyeline. Same content, same control, same one-per-screen rule.
The fourth state, and the one the three-tab screen has no room for. An open conflict about the contract sum sits at the root of the projection, so no reading is offered at all — a wrong verdict here is worse than none, and saying so is a sentence rather than a color. The route out is one tap and it belongs to the homeowner, which is why it is a real button.
Everything on 7b that cannot yet be served is absent rather than zeroed, and the screen is composed around the two figures that can. Where the projection goes is a container-shaped absence at the outline rung — no fill, no lift, because there is no object to raise — and it carries the one action that ends the wait. The verdict line reads "not enough to say", which is a legitimate answer and not an apology.
The earlier explorations are gone and what survived is one screen with three views. Health is the four vitals, the categories split into settled and still moving, and phases against typical. Trend is the forecast line replotted at every bill and quote, with the three readings stacked under it — the view that shows what HouseChalk has been doing in the background for fourteen months. Where you are is the milestone timeline, reached behind you and windows ahead. 7a is the phone, with a segmented switch at the top: on a 390px column the three views are too tall to stack, and tabs beat a very long scroll because each view answers a different question. 7b is the desk, where the switch disappears and all three sit in one reading order, forecast across the top.
The switch is live — tap between Health, Trend and Where you are. The header above it stays put, so the one-line verdict never leaves the screen. "Mark one of these finished", under the still-moving run, is live too: it opens the sheet that closes a category out by hand. That is the route for a homeowner with no builder connected and no draws, where nothing exists to infer a finish from — the queue proposal in Record is the shortcut, and this is the door.
This is the earliest a line can be seen going over, and it is the reason the projection moved this week. It stays inside your contingency, which is why the build still reads Steady.
Two steps in fourteen months: the rock under the slab in June last year, and the HVAC quote on Thursday. Between them the forecast sat flat for eleven months, and it has stayed under the contingency ceiling the whole way.
The step is the truss delay. It stopped growing the week the frame went up and every milestone since has landed in its own window.
Two change orders in fourteen months, both drawn from the finishes allowance, so the contract figure has not moved. The two steps in the forecast above are quotes coming in over their lines, which is a different thing from a change order.
Twenty-two answered, one open. The open one is the cabinet change order, and six to ten weeks of shop time start on your signature.
A quote landing over its line is the earliest warning the forecast gets, months before a bill would show it.
Six to ten weeks of shop time start when you sign, and the counters are measured off the cabinets once they are in. A week spent deciding is a week added to the two windows above.
With the width available the tabs are unnecessary and the reading order does the work instead: the verdict and four vitals, the forecast over time, then the two ways of digging in side by side, with the timeline and the one thing worth watching at the foot.