| <!-- SPDX-License-Identifier: Apache-2.0 --> |
| <!-- Copyright 2026 alvations (Melon Lab) --> |
|
|
| # Critical review, fifth pass (beta-1) |
|
|
| The same deliberately harsh but impartial reviewer, fifth visit, judging the |
| beta-1 build. `docs/REVIEW.md` (6), `REVIEW2.md` (7), `REVIEW3.md` (7.5), |
| `REVIEW4.md` (8) are the record and stay untouched. The headline this round is |
| **difficulty**: an easy/normal/hard setting, a hard mode that finally escalates |
| the loop *in kind* (double changes, persistence, reverts), an age-match reading |
| register for younger players, and a coach-only trail that earns the false exit. |
| For four reviews I have said the polish was all on the edges and the middle never |
| changed. So the only question that matters: did the middle finally move. |
|
|
| **Verdict: 8.5/10, up half a point.** It did. Hard mode is the first change in |
| this whole series that makes level 8 a genuinely different task than level 2, |
| which is the note I have hit every single time. Easy mode and the reading |
| register are real accessibility, not theatre. The coach's passenger now leads you |
| to its secret instead of hiding it behind a dice roll. This is the build where |
| the game stopped only decorating the room and started rebuilding it. It is 8.5 |
| and not 9 because the depth it just built is opt-in, hidden behind one win, and |
| absent from the setting most players will actually sit in, and because the arc |
| count, the true ceiling, is still three. |
|
|
| ## What actually improved (and it is the core, for once) |
|
|
| - **The loop finally escalates in kind.** Two simultaneous changes, an anomaly |
| that persists, a clean revert: these are new *problems*, not a heavier coin. |
| This is the single thing I asked for in four reviews and it is here, built |
| correctly, with the fairness guard intact and no answer leak. Credit where it |
| is very much due. |
| - **Difficulty is a real accessibility win.** Easy (fewer changes, a floor so it |
| is never trivially "always go on") plus a plain-language reading register means |
| a child or a nervous newcomer can meet the mechanic on their terms. That is the |
| right instinct and it is done without dumbing the real game down. |
| - **The coach trail is the best kind of fix.** The false exit was my "your best |
| content is a 2% lottery" complaint made mechanical; now the passenger has to |
| clock you, says its one line, and arms the exit. Diegetic, earned, discoverable. |
| - **Settings travel with the save.** A small, correct touch: difficulty, reading, |
| display, and language ride in the export now, so the setup is portable, not |
| just the progress. |
|
|
| ## The criticisms |
|
|
| ### 1. You built your best mechanic and then hid it |
| Hard mode answers the deepest complaint in this series and it is locked until the |
| first escape and off by default before that. The first full playthrough, the one |
| that decides retention, is still the shallow single-dial game. I understand the |
| progression logic (teach the base skill, then deepen). But the effect is that the |
| version of the game most likely to make someone stay is the one they only reach |
| after they have already decided to. Let people *choose* hard. |
|
|
| ### 2. Three arcs. Still. Five reviews. |
| This is now the oldest unaddressed note and it has to be the headline criticism. |
| Every pass has added systems (sharing, Act 2, the collection, records, and now |
| difficulty) to the *same three places*. Difficulty multiplies the ways to play |
| three arcs; it does not add a fourth. The engine is begging for more world and it |
| keeps being fed more machinery for the world it has. A fourth and fifth arc would |
| do more for beta-1 than any system added since REVIEW2. |
|
|
| ### 3. The surface area is getting genuinely large for a minimalist game |
| Tally the opt-in layers now: sharing, a 20-badge board, a fragment codex, a |
| records ledger, promo codes, a settings panel with text/brightness/motion/ |
| difficulty/reading, save-load. Every one is defensible; together they are a lot of |
| scaffolding around a two-minute loop, and the aesthetic argument of the whole |
| thing is restraint. The game is quietly becoming a menu. Watch that the chrome |
| does not outgrow the room it frames. |
|
|
| ### 4. Difficulty settings are utilitarian intrusions on a careful mood |
| "Easy / Normal / Hard" and "Simple / Normal" are the least in-world words in the |
| entire experience. On a landing that spent enormous care making everything feel |
| like a place and not a product, the settings panel now reads like an options |
| menu. Fine, functionally. But there is a version of this where difficulty is |
| named in the game's own voice, and the current one spends none of that voice. |
|
|
| ### 5. Hard's double-change can't teach itself |
| A two-change loop is only a lesson if the player can learn "keep scanning". They |
| can't: turning back on either change is correct and ends the loop, so a player |
| never finds out a second thing had moved. The mechanic is sound and the pedagogy |
| is missing; the skill it wants to build has no feedback path. |
|
|
| ### 6. Normal, where most people live, is unchanged |
| The defaulting sends everyone to hard after one win, but the moment hard is too |
| much they land on normal, which got none of the new kinds. So the setting with |
| the largest population is still the meta-gameable base rate from REVIEW2. The new |
| depth is only at the poles. |
|
|
| ### 7. The coach's arming utterance is a tonal outlier |
| Mechanically the trail is right. But the passenger's fixed line is the most |
| explicit, expository sentence in a game whose whole voice is implication. It |
| tells where every other line suggests. It earns the exit; it also briefly breaks |
| the spell. (This is a prose note; the aesthetics review will have more.) |
|
|
| ## If you fix only five things, in order |
|
|
| 1. **Let players choose hard from the start.** Stop gating your best mechanic |
| behind beating your weakest. |
| 2. **Build the fourth arc.** The five-review-old note. World, not systems. |
| 3. **Give normal one new kind.** Lift the setting most people sit in out of pure |
| base-rate; the revert alone would do it. |
| 4. **Teach the double change.** A post-hoc tell so "keep scanning" can be learned. |
| 5. **Name difficulty in the game's voice.** Or hide the panel deeper. Do not let |
| an options menu be the least atmospheric screen in an atmosphere-first game. |
|
|
| ## Pull quote |
|
|
| > Five visits in, the middle finally moved. Hard mode is the deepening I kept |
| > asking for, and it is real. Now the game has one more problem than it used to |
| > and exactly as many places to solve it in as it had at review two. Stop hiding |
| > the depth you built, stop adding systems to three rooms, and build the fourth |
| > room. |
|
|