Somebody arrives at the map having already saved something. The page hands them a search form. Four ways to fix that, drawn to scale against the fold, with the trade each one makes written beside it.
Every number on this page is computed at load from one table of measurements taken in headless Chromium against the built page — including the height of the row Option C proposes, which is read off the row this page actually draws. Nothing here is typed in twice.
Measured against the built index.html in headless
Chromium at 390×844, 768×1024 and 1440×900, Show 5, after a
default Near me search — the state one tap of Find 'Em leaves you in.
1280×800 came back identical to 1440 below the gutter, so wide is one
case rather than two.
| Landmark | Phone 390×844 | Tablet 768×1024 | Wide 1440×900 |
|---|
The page was measured twice at each size: once with empty storage, once with three bookmarks and a saved Austin → Dallas itinerary.
| Landmark | Nothing saved | 3 bookmarks + 1 itinerary | Difference |
|---|
The one difference in the whole document is the text inside
#yerStuffSum — “3 bookmarks · 1 itinerary”
instead of “Nothing saved yet” — inside a section that arrives
closed, at the bottom of the page. There is a route to it: below 700px
the section links move behind the burger, so it is two taps. Nothing below
proposes changing that route. What is missing is that nothing above the fold
ever suggests taking it.
Drawn at the real viewport width and scaled to fit, with the dashed line where that screen's fold is and everything below it dimmed. Switch This visitor has at the top between Nothing saved and 3 bookmarks and watch which options change and which do not — that is most of the argument. Switch Viewport to Wide 1440 and the frames redraw as the two columns that screen actually gets: the map on the left, the results on the right.
Real controls, not pictures of them — .seg,
.act, .cbx and .findem are lifted from
index.html as they stand. Press things.
Near me | Roadtrip | Yer Stuff. One line of markup, and
.seg is already how this page says pick exactly one of
these. Press the third one, then press Find 'Em.
Texas Find 'Em
Rejected, and it is a category error worth writing down. The other two options in that group are search modes: they change what the Find 'Em button searches over. Yer Stuff is not a search mode — it changes what the page is, and pressing the primary button while it is selected has no defined meaning. A segmented control whose third option makes the page's main button meaningless is carrying two ideas in one row. And it puts retrieval inside the search form, which answers the wrong question.
A is only broken because the box underneath it did not change. So change the
box, and then follow it all the way through. Search mode becomes
MODE with three intents — Near me, Roadtrip,
Saved — and picking one swaps everything inside the panel,
including the primary button. Saved loads what is in Yer Stuff today,
and the Yer Stuff section is then deleted: out of the page, and out of
the nav. Click through all three, then flip the visitor to
Nothing saved.
Texas Find 'Em
It is the only option that makes the page smaller. Every other one on this bench adds something. This deletes a section and a nav link: The segment already exists, so the mode itself costs nothing, and Find 'Em stops being orphaned — it belongs to the box, and in Saved there is no box for it to belong to.
And it answers the empty-room problem that sinks B, by accident. Settings and Start over live in Yer Stuff's third tab today. Delete the section and they have to go somewhere, and the only somewhere is Saved — which means Saved can never be gated on having bookmarks the way C's row is. That sounds like a loss and is not: a first visitor who opens Saved does not find an empty room, they find their settings. B's tab is empty on a first visit; this one never is.
What it costs, and it is the largest change here.
It is a mode, and modes hide: in Saved the search form is gone and
getting back is a deliberate act, on a site whose main job is search. A
primary button that is absent one time in three is something the page has to
teach. It deletes a section that has a fragment — #yerstuff is a
URL people have been given and the app's own nav links at it — so that
fragment needs somewhere to land rather than nowhere.
And it is not compatible with C: C's row is a shortcut
to a section, and this option removes the section. These two are rivals, not
halves of one design.
E, followed the rest of the way. The box gets four tabs across the top — Search, Places, Roadtrips, Yer Stuff — and the three sections that used to sit under the map become the last three of them. Only How It Werks stays in the nav. The map and the results do not move: the tabs are on the box, not above the page, which is the whole difference between this and B. Click the tabs.
Texas Find 'Em
It takes the most off the page of anything here. And it is the only option that makes the site's structure simpler rather than adding to it: one box that shows one of four things, a map, results, and a single link in the nav.
It also answers the Roadtrips question in section 6 for free. If the drive cards are a tab rather than a fold at the bottom of the page, then Find 'Em on a card has somewhere obvious to go: it switches the box to Search in Roadtrip mode with A and B already filled. Pick a drive, and the box you picked it in becomes the box that searches it. That is one gesture instead of scroll-open-read-scroll-back.
What it costs, and it is the biggest bill on the
bench. Four tabs is a lot to put above a form that is already 823px on a
phone, and three of the four are things somebody is not usually there for.
Three fragments — #yerstuff, #theplaces,
#roadtrips — are URLs the app's own nav and other pages link at,
and they all need somewhere to land. Every one of the four tabs has to fit a
390px row, which at four is tighter than the three-up segment the panel uses
today. And it is a mode four ways: whatever you are
looking at, three of the site's four rooms are not on screen.
F with two tabs instead of four, and the difference is not the count — it is that Yer Stuff stops existing as a pool. Places and Roadtrips are the two tabs, and each one carries two modes: a search, and the things you already saved of that kind. A bookmark is a place, so it lives under Places. An itinerary is a trip, so it lives under Roadtrips. Click both tabs and both modes.
Texas Find 'Em
It fixes the one thing that would have sunk F. F's four tabs measure 69–73px each at 390 — they fit, with no slack, and a fifth room or a longer word than Roadtrips breaks the row. Same page arithmetic as F otherwise: the same tab row above the panel, the same three sections gone below it.
And the information model is better than any of the others, including F's. Every option above this one treats the stuff you saved as one room. It is not one thing — it is bookmarks and itineraries, which have nothing in common except that you kept them. Filing each next to the search that finds more of its kind is the first arrangement here that a visitor could have predicted. It also promotes Roadtrip from a mode buried in a search panel to one of the site's two halves, which is how the rest of the site already talks about roadtrips — they have their own pages and their own sitemap.
The tap count, carefully, because it is the axis this whole project turns on. The page opens on Places · Search, so the Yer Places control is already on screen: bookmarks are one tap, the same as F's Yer Stuff tab. Itineraries are two — Roadtrips, then Yer Trips — where F has them at one. So G is level with F on the common case and one tap behind on the rarer object, against today's two taps and 2,957px of scroll for either.
Two things to settle before this could ship. Places is doing a new job as a word — today The Places is the chain roster, and a tab called Places that opens a search form will read wrong to anybody who knows the old one. And the mode labels want to be parallel: Search / Yer Places against Searches / Yer Trips is one plural out of step.
Two tabs at the very top: Find something and Yer Stuff. The search panel, the results and the map live under the first; bookmarks, itineraries and settings under the second. Press Yer Stuff, then flip the visitor at the top of this page to Nothing saved.
Texas Find 'Em
Kept on the bench, and it loses on cost. It hides
the map behind a tab, on a site named for a map. It makes an empty room a
first-class destination on every arrival, including every first one — and
first-run is the majority of visits. And it is a large structural change:
#yerstuff leaves the page flow, so the three jump links, the
sticky bar's “the one you are in” rule,
openFold(), the #yerstuff fragment and every shared
link to it all change meaning in the same commit. If this
site ever grows accounts, B is the shape it becomes. It is not the shape
for today.
One row above the search panel, holding the three most recently marked bookmarks and the most recent itinerary. Tapping one selects that stop and lights its dot rather than scrolling you to a section where you can then tap it. With empty storage it renders nothing at all — not a placeholder, not an empty state. Open it, press a row, then flip the visitor to Nothing saved. Then switch Row shape and watch what the always-open version costs.
Texas Find 'Em
Recommended. It is honest about the asymmetry —
retrieval is only a need for somebody who has something to retrieve, so it
charges the cost only to the arrivals that have it, the way Coming soon
and Open now are already hidden until there is something
to filter. It writes nothing new: setMark() already stores
{ t } on every bookmark and itinerary and renderStuff()
already sorts by it, so “the three most recent” is a read of data
that is already there — no new dated signal, which is a door
CLAUDE.md closed on purpose. And it says nothing new about
anybody: an itinerary's two town names are exactly what the
CITIES-only rule was built to keep it down to.
The one real cost. A second surface rendering the same records — which is how this repo has drifted before. So the row has to be built by the same function that builds a row in Yer Stuff, in its compact form, with a check asserting it rather than a comment asking for it. If the row gets its own renderer, this option becomes a liability instead of a feature.
No new interface at all. The page already picks an intent three ways — a
?chain= hand-off wins outright, otherwise the most recently
touched draft comes back in front of you with the search re-run, otherwise
Near me. Add a fourth branch: if the last thing you touched was an itinerary,
open on it.
Rejected on two counts. It cannot serve “I want the other bookmark” — a restore only ever offers one answer, and retrieval is a browse. And doing more of this silently is what makes a page feel like it has opinions about you: the draft restore gets away with it because it announces itself in the origin line and is undoable in one tap. Guessing further, with no control to disagree with the guess, is a worse version of the problem this whole bench is about.
The row is real and its height is measured off the one drawn above, so this arithmetic moves if the design does.
| Landmark | Today | With the row | Change |
|---|
The cost lands on the search path and is stated rather than designed around: somebody with saved stuff gets it above the fold and pays with the Find 'Em button moving further from them. Somebody with nothing saved pays nothing, because there is no row. This is also why the row is capped at four and does not scroll — an unbounded list would push the search panel down in proportion to how much somebody has saved, which punishes the people who use the site most.
Order matters. Mobile Optimization moves the same pixels the other way: its Option 1 collapses the search panel after a search and takes Find 'Em from 977 to 565. These two should be measured together, and this one should land second, so its cost is measured against the collapsed panel rather than the current one. Ship this first and both proposals end up quoting figures that were true on different days.
Same project, same idea: the fastest thing a page can do for
somebody is hand them a prepared thing instead of a form to fill in. Today the
Roadtrips fold on the home page holds three category cards, and the only
way into an actual drive is to leave for /roadtrips/ and come back.
Proposed: the six drives across the Texas Triangle, each with a button that
runs it.
Three things change and one of them is the point. Find 'Em is new: it loads that drive in Roadtrip mode with both ends filled and runs it, so picking a drive and searching it is one press instead of a trip to another page and back. More keeps doing what it does, going to that drive's own page. And the gray line counts stops rather than drives — Every one of those numbers is the number that drive's own page already prints, read off it rather than recomputed here, because a card whose count disagrees with its destination is worse than a card with no count.
What it costs, and it should be decided rather than absorbed. Fourteen drives — every Major route and every Short trip — lose their entry point on the home page. Dallas–Waco, Houston–Corpus Christi and San Antonio–Laredo are only reachable by going to the roadtrips page, which is one more link than they have today. The trade is that six real drives beat three abstractions, and the Triangle is where most of the driving is. A seventh card reading All 20 drives → would close that hole for the price of one card, and is the obvious thing to try before shipping this.
And it wants a home. Six cards is more than three, so this fold gets taller — which is a cost on the phone page that section 5 is already counting pixels for. If Option F lands, this grid is the Roadtrips tab rather than a fold at the bottom of the page, the cards are what that tab holds, and Find 'Em on one of them switches the box to Search in Roadtrip mode. That is the version where these two changes are one change.
Nothing stamps this file and nothing verifies its contents. If
index.html moves, this bench goes stale in silence — which is
acceptable for a bench, and is the reason to delete it once this ships. The
private argument, with the losing options and the open questions, is
PROPOSAL.md beside it.