Plan option

Anatomy

Five slots and four atoms, which makes it the densest molecule in the round. Everything below is drawn by design/system/components/plan-option.css and the atom files it hosts, through the same entry point a product screen loads. Three of the sizes you can see here are not in this file, and that is the point: the badge is chip.css, the price is amount.css, the "Everything in Tendd Pro" label is eyebrow.css, and this file only says where they go.

Synced from the product, 2026-09-30 (P26, D4 amended). The founder, on Upgrade from inside the product: "очень плохо выглядит выбор планов". Measured then at 1280: the same list printed three times, so 60% of each card was a repeat and the three prices were the smallest thing in a wall of ticks, and the buttons stood in a row under the cards, so a price was read in one place and pressed for in another. Since 2026-09-29 each card carries its own button at its foot, the yearly card is marked with the selection edge (.plan-opt.best), the cadence sits beside the figure on one baseline, and what Pro includes is said once, in one .incl under the row, in two columns past 760. The row is three columns from container 900 and stacked below (Grid). Prices are D4 as amended on 2026-09-30: $9.99 a month, $69 a year. No string on the card changed except the saving, which the new monthly price moved from $27 to $50.

The stage is past 900 at a 1280 window, so the row stands in three and the list runs in two columns; narrow the window and the cards stack, one full-width card each, which is the phone's arrangement. In the product the row is a form wearing .plans, because each card's button posts its interval (P5); here it is a div and the buttons are links, which the grid does not tell apart.

• .plan-opt the molecule: a div in a flex column, with the container hairline and the container corner. .best, since 2026-09-29 (P26), gives the yearly card --line-selected as its edge: petrol's "current selection" job, and the landing's same card. NOT a control, which is why it keeps the hairline while the door keeps the control edge
• .chip.tone-accent the chip atom in its accent tone, on one of the three cards. It was .best until step 6. It leaves the flow on both hosts since 2026-08-18: this file takes it out of the column and drops its margin, and each host says where it goes - the top edge on the landing, the top right corner here. The position: relative this file has carried on the card since stage 07 is the anchor both of them use
• .amt the amount atom at its 32px plan size, declared in amount.css. The card's one big thing
• .per the cadence, "a year" or "a month", and it is this card's own slot, display: inline beside the figure since 2026-09-29, on one baseline: it lives INSIDE the amount and the amount atom does not declare it. It resets the weight, the tracking and the ink so "$69 a year" does not read as one 32px phrase
• .p the sentence about what the price means. flex: 1 1 auto, so it absorbs the height difference between three cards of unequal copy. It is what keeps the three actions and the three includes lists on one line, measured at 360 and at 1100
• .btn the button atom, stretched to the card. One filled, one outline, and the third card has none. It is the one button in the product allowed to wrap, since 2026-08-13: button.css gives .plan-opt .btn a white-space: normal that no other button gets, because this label is a sentence carrying a price and a period. Measured at 390px: "Start Tendd Pro - $7.99 a month" is 249px at the browser default and 356px at a 24px root, in a 358px track, and the page scrolled sideways by 16px before the exception. The rule lives in button.css and not here, which is this file's own header holding: white-space is the button's property and a host reaching in to change it is the defect
• .q what stands where the button would be on the card that has none. It reads --tap as a geometric FLOOR rather than as a target: it stops the slot collapsing below a button's height when its sentence is short. It does not do the aligning, .p does
• .incl what the plan includes, under a divider, below the action and never above it: the price and the button are the decision, the list is the reassurance a person reads after making it. Since 2026-09-29 (P26) it is outside the cards, one block after .plans, said once instead of three times; its selectors lost their .plan-opt scope and its label's placement moved with it in eyebrow.css. Six rows now: "Every name in your bank scan" and "A bank connection that keeps the list up to date" joined when the bank became Pro (D-Gate)

Variants and sizes

One axis, and it produces no visual difference at all: the card is declared once for two hosts. What looks like a second axis, "with a button" against "with a question line", is not a variant of the card but a fact about the plan, and it is in the second table because a person reading this page will look for it.

The price is 32px and not 46px, and the gap is the One Number Rule. 46px belongs to the monthly total and to nothing else in the product; a plan price at that size on a screen that is not about the person's own money would be the loudest thing they have ever seen in Tendd, on the screen where they trust it least.

When to use it

A person meets three of these on Upgrade and on the welcome page, and nowhere else. (Upgrade only, in the product since P16: the landing's pricing is its own organism.) The card exists because a price is a comparison: three prices in three identical frames is the only shape in which $69 a year and $9.99 a month can be read against each other, and D4 fixes both numbers (the monthly one moved from $7.99 on 2026-09-30, the yearly one did not). The screen sits at depth and never in front of basic visibility, which is D3: the list, the total and the alerts are free and uncapped, and this card sells history, trends, advanced alerts, cancel guides and the analytical export, and since D-Gate (2026-09-02) every name in a bank scan and a bank connection that keeps reading.

Its copy does the arithmetic for the person, from voice/docs/microcopy.md: "That is $5.75 a month, and it saves about $50 a year versus paying monthly." (It said $27 until 2026-09-30, when D4 was amended to $9.99 a month; microcopy's rewrite log carries the arithmetic.) The button carries the amount as well as the verb ("Start Tendd Pro - $69 a year"), because nobody should learn what they are being charged after the tap.

The rule, and the anti-rule

Limits

U12, context. A state change is colour over --dur-state, and a component declares a transition only where its host is not a native interactive element. base.css puts the one transition on :where(a, button, input, select, textarea, summary, label) and moves four properties: background-color, border-color, color, text-decoration-color. Nothing on that list moves an object, and the focus ring is deliberately off it. This component is one of the three the rule names, and it declares the transition itself because its host is a div: border-color alone, reading the same two tokens so the two can never drift. Full table with sources and how to check: Usage rules, and the register is on Motion.

2026-08-18: three prices that did not start on one line

The founder opened the coloured Upgrade screen and said “выглядит не очень”. The loudest thing in that screenshot is the comparison failing at the first glance: only the yearly card carries a badge, the badge was a block in the flow above the price, so $69 stood 32px below $7.99 and Lifetime - measured at every width from 768 to 1920, on the one screen in the product whose whole job is a comparison between three numbers.

The answer had been written three days earlier, inside one host. The landing plan hit the same defect on 2026-08-15 and recorded all three arrangements it tried: “Stacked above the price it pushed the yearly figure 31px below the other two, so the numbers a person compares did not start on one line. Beside the price, in a second column, it fixed that and cost the card its axis... On the top EDGE it costs no flow height at all, so the price is the first thing in every card.” It even quoted this file's position: relative and called it “the day” the anchor was needed. That fix went into .landing .lp-plan, so the app's card kept the defect - the same shape as the story's release the day before: a correction written in one arrangement instead of in the component.

So the flow height is the component's rule and the place is the host's. The chip is out of the column here with margin: 0, which is what makes the amount the first box on every card and the three prices share a line by construction; the landing keeps its straddle and no longer has to undo a margin this file set. Top right and not the top edge on this host, on the axis argument the landing named: a centred card has no corner to belong to, and a left-aligned card standing shoulder to shoulder with two others would read a straddling badge as a label for the row. Measured so the corner is not a guess: the badge is 66 wide, $69 is 60.41, and the narrowest card content this screen has ever had is 201 - 74.6px of clear air at the tightest width in the product, more at every other. Verified: the price rows are 0px apart at every three-column width from 760 to 1920.

And the card finally got the width its own button was measured for. The button label wrapped to two lines at every width from 768 up, with “year” alone on the second line - and the cap it wrapped inside was widened at stage 07 because the button label wrapped. The arithmetic behind that widening had answered the eyebrow (181.23 needed, 201 given) and not the button (248.78 needed, 201 given). The grid carries the recomputation and the band below a 913px window where the label still wraps; here the consequence is that the card is 293px instead of 235 past that window, so the supporting sentence reads in three lines rather than laddering into four of three words each. The band that is still wrapping is a row in docs/backlog.md, open and addressed to the founder with Voice for the string.

States

One state, and it is the open question of this round. The card is not a control: it is a div holding a button, it cannot be focused, it cannot be pressed and it has no accessible role, which is exactly why the consolidation refused to treat it as one and why it keeps the container hairline. The four states a person actually uses on this screen belong to the button inside it and are on button.html.

And since 2026-08-16 that one state FADES. The card is a div in all six of its placements, three on the landing and three on upgrade.html, so it falls outside the element list base.css carries the product's one transition on - by exactly the same argument the paragraph above makes: it is not a control. It declares the transition itself, on border-color alone, because that is the only thing its one state moves. The rule and its tokens are in Limits above.

IT HAD A POINTER RESPONSE UNTIL 2026-08-20 AND THE FOUNDER DROPPED IT. kit.css strengthened the card's edge on hover, from the hairline to a darker hairline; the value had no role, so stage 08 declared one, and the question underneath stayed open for eight days: should a card that cannot be clicked answer a pointer at all? It is answered no. The card is a div holding a button - no focus, no press, no accessible role - and an edge that strengthens under the pointer is this product promising a click it does not deliver. The button inside keeps its own state and is the thing that was always meant to be pressed.

THREE DELETIONS AND ONE DECISION, which is the part worth keeping. The hover went, so --line-container-hover had no reader anywhere in the system, so --hairline-strong and its computed dark pair had no role. Each step follows from the one above it and only the first was a choice. Measured before removal, on Upgrade at 1280: the card's edge went rgb(228, 233, 234) to rgb(205, 215, 217) under the pointer and nothing else on the card moved. That is the whole pixel cost, on three cards on one screen. The two answers, kept for the record, and the second one won:

Rest, live

Hover the card and then hover the button inside it. The button responds and the card does not, and since 2026-08-20 that is the answer rather than the question: the thing that can be pressed says so, and the thing that cannot stays still.

Rest, in both themes

The states are described, not photographed. Founder's decision, 2026-08-11: the snapshot strip went from every page in this kit, 120 images across 18 pages. They were framed as measurements rather than illustrations, and that framing was honest, but the bill came due all session: a value moves, and every picture of it is quietly stale until somebody re-shoots it. A stale picture is worse than no picture, because it looks checked. What is left is what a picture never carried anyway: the state, and the token it reads. The resting component is live and interactive above, so hover it, tab to it, and read the value off the real thing rather than off a photograph of it.

The technical half

The second carried question, said plainly. .plan-opt .p paints the muted line atom in body ink. That is a host restyling an atom, which is the defect this round exists to remove, and it is carried rather than fixed for two reasons: dropping it changes three paragraphs on a live coloured screen with no decision behind the change, and the fix is not this file's to pick. The two answers are an emphasis modifier declared on the muted line, or this card accepting muted ink the way every other card in the product does. Recorded here, in the component's own file, and in the report of this round.

Lives in design/system/components/plan-option.css. Stands on 2 grey pages and 1 coloured one, Upgrade. The landing's three cards are still grey, so that host's render is read from the grey. In the product on 2026-09-30 it stands on Upgrade only: the landing sells with .lp-plan since P16. The grid around it (.plans, one column and then three) is the Grid organism.