Bunkers that are actually different from each other (§7)

[0xward]

§7's last [TARGET] flips to [TODAY]. No smart contract, ABI, or contract address is touched.

What was actually there before

Not "the scaling half-exists already", which is what §7 claimed. Measured with the new npm run measure:acts, which mounts the real engine at every act:

Bunker 5 vs Bunker 1
walkable floor area	1.00×
rooms	0.95×
enemies	0.93×

The last bunker was very slightly emptier than the first.

Floor size came from depth alone, which runs 1–5 inside every bunker, so the act never entered the arithmetic. Worse: actHardMode had one entry, for act 4 — so bunkers 1 through 4 were mechanically identical, since enemy stats also scale on depth. The only thing that changed across those four was which archetypes could roll: variety, not difficulty.

"Which bunker should I raid?" had no mechanical answer to give.

The two axes, now real

Risk — actHardMode covers all five acts. Gentle at the start so a new player's first raid is not harder than the campaign taught them; steep at the end so the last bunker earns its T3 shards. Act 4 is unchanged at 2.0/1.3 — those numbers were tuned against the boss one-shotting players, and nothing here reopens that.

Time — ACT_SIZE in dungeon.js adds width, height and one room per act. Rooms are what carry enemies and containers, so effort and reward scale together without a second knob.

Bunker	Risk	Enemy HP	Shards	Length
1 · Treeline	▰▱▱▱▱	×1.00	T1	—
2 · Sunken Field	▰▰▱▱▱	×1.20	T1	+15%
3 · Frostline	▰▰▰▱▱	×1.45	T2	+30%
4 · Hollow Market	▰▰▰▰▱	×1.70	T2	+55%
5 · The Last Light	▰▰▰▰▰	×2.00	T3	+65%

Measured after: 1.65× the area, 1.55× the enemies. Deliberately modest — a MiniPay session is 3–10 minutes on a mid-range phone, so bunker 1 stays the short option. That is the point of an axis rather than a global difficulty bump.

And the player can see it

A difference nobody is shown is not a choice. The map's bottom bar now carries risk pips, shard tier and length under the bunker's name — on the way to the ENTER button, not somewhere to go looking for. Locked bunkers show it too: knowing Bunker 5 is the only T3 source is a reason to keep going.

Keeping this file honest

lib/constants/bunkers.ts is the same shape as the bunkers block that made game-config.ts a liability — a table describing bunkers, sitting next to the engine that really decides. So it is guarded differently:

npm run test:bunkers mounts the engine, reads its own actHardMode and shard-tier rule, measures floor size per act, and fails if any number in the constants disagrees. The engine stays the source of truth; the constants are a view of it the React shell can import (the engine is a static <script> — rule 1).

A bug this surfaced

apply() in WorldMapHub runs twice — once from the instant localStorage draft, again when the Firestore copy lands, typically a network round trip later, which is exactly while the player is looking at the map. Its clamp (prev > highest ? highest : prev) ran both times.

So tapping a locked bunker to read what it holds snapped the selection back to your own bunker a beat later, with no explanation. The clamp now applies only until the player taps something.

Found by the UI test failing on an assertion I had assumed was about my own markup — I spent three rounds treating it as a test race before dumping the panel HTML and seeing the selection had genuinely reverted.

One presentation fix

Length reads "+65% LONGER" rather than "1.6×": at one decimal place bunker 4 (1.55) and bunker 5 (1.65) both rendered ~1.6×, so two rungs of the ladder looked identical on the one line meant to tell them apart. A test now asserts every note is distinct.

Verification

Check	Result
npm run measure:acts (new)	the before/after table above
npm run test:bunkers (new)	34/34 — constants checked against the live engine
npm run test:bunkers-ui (new)	15/15 in a real browser, three consecutive runs
npm run test:traversable	0 failures in 100 floors — bigger floors stay walkable
npm run test:seeded	15/15
npm run measure:bunker	fragment thresholds unchanged (act 0 is untouched)
npm run test:raid · test:contracts · test:floors · test:inventory · test:streak · test:streak-ui	all pass
tsc --noEmit	clean
lint	34 errors — unchanged baseline
audit · check:copy · check:market · check:cssvars · check:attribution	all pass
build	succeeds

Docs: GAME-DESIGN.md §7 rewritten with the measured before/after, §2 marker and the build order updated, and game-mechanics.md's bunker table now carries the real numbers instead of "Beginner / Easy / Medium / Hard".

Still open

How long is one bunker in real play? The sandbox can measure rooms and enemies, not a person's minutes. If a full 5-floor clear of Bunker 5 runs past ~10 minutes, a "raid one floor and extract" option is the next thing to add — that is the shape a MiniPay session actually has. Left as a stated open question in §7 rather than guessed at.

---

Generated by Claude Code

[chatgpt-codex-connector]

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
nullstate	Ready	Preview	Jul 31, 2026 6:57am