The first diversityPossible() call after a save loads (migrate() checks every
customer's requirements) built a graph of every buildable link between all
~28k city pairs, scanning all 475 cable segments for each pair, with string ids
and objects for each of ~19k links: about 0.8 s on a fast machine, ~0.9 s of a
throttled phone's boot. Now the cables come from a table per pair and the graph
holds city indices and one number per risk (each land trench and own cable on
its own; every segment of a cable, or a fabric, sharing one). The answers match
disjointRoutes() over the old link list for all 14,042 sampled pairs, won or not
and with scenario bans.
bezGeom() measured every city against every candidate curve, for every link the
first time the map drew it. A curve stays inside the box of its end and control
points, so a city further than the clearance it needs outside that box always
measures as clear; leaving it out gives the same curve. About 4x faster, and
the curves on the big fixture are byte-identical.
The sea waves, hills and the far view's islet list test points against every
coastline polygon while the page loads: about 0.8 s of a throttled phone's load.
A point outside a ring's box is outside the ring, so nearly every test stops
there. Waves, hills and the far view's islands come out byte-identical.
dayTick() routes the day, then appends today's transit use to t.hist. Every
compute() after that (any click, the intern, the sales trial) took the last 29
entries of t.hist, which now end on today, and added today's use again: today
counted twice and the 30th day back dropped out. So compute() gave a different
transit bill the second time after each day. p95() now leaves today's sample out
once S.hist says the day has been recorded.
This was the only part of compute() that was not a pure function of S. The rest
of the drift round 1 saw (R right after dayTick vs. a fresh compute) is dayTick
itself changing S after its compute (customers churning, events ending), which
is intended: the next day routes the new state. Scenic circuits' hysteresis is
documented, with a test that two calls converge.
After three stars, Next puzzle is the one big button (and has the focus); the
results card counts the stars still to win in the pack; and when one more link
within budget would earn the third star, the card and the results say so.
On phones the puzzle card's Undo, Reset, Back and menu buttons are 40 px touch
targets and stay pinned at the foot of the sheet when longer languages make it
scroll. Goal cities show a leading mark (○ to reach, ◐ POP open, ✓ joined)
instead of a trailing one that looked cut off.
Toasts are cleared when a puzzle starts and when you leave it, and every
main-game action (ops, inbox, bank, building, routers and the rest) does
nothing while a puzzle is on; only the puzzle's own actions, menus and
settings work there.
Progress is recorded on every change, not only when the stars go up: more stars
start the best spend again, and a cheaper build at the same stars replaces it.
The first pack leaves 16-18% over the cheapest three-star build, the Pacific and
speed packs about 10%, and the N-1 Club 6-7%. The first pack no longer promises
no budget. A brute-force check in the tests keeps each reference build the cheapest.
The puzzle price and the city panel now use the city's own monthly POP rent
(Suva is 1.5x), not the plain one, matching the higher setup already charged.
Two waves on the same cable now share one failure, as in the main game's
diversity checks. Kiwi Fruit Salad gets Los Angeles as a stepping stone (Seattle's
only cable south is Hawaiki), Samba Sprint's budget rises to $1.06M, and the
weekly reference ring never puts two hops on one cable.