v0.3.5: the launch version

The version that goes to launch, and the first with an outside name in it.

• First community contribution merged: UI text onto the strings map, thanks to @Aayush10016, hours after the repo went public
• The in-game rulebook now matches the engine exactly (trust erosion under unrest had a threshold the code never had)
• The deploy script and a registry link checker are committed: the last mile is as inspectable as the rest
• The design constitution gained the constants appendix its citations resolve to
• Balance docs point at the CI invariants that guard the thesis
• Commit timestamps are batch-normalized before publishing; governance now says so up front
• security.txt, wider landing-page count guard, METR quoted as printed, founding decisions annotated where amended

Play at https://criticalwindow.org. Paper kit attached.

v0.3.4: checkability holds

Everything a skeptic checks first now checks out.

• Every commit hash cited in the balance history resolves on the published branch, enforced at deploy time
• The evidence map's worked example and its table quote the same METR figure
• Pedagogy research carries the evidence class its claims deserve; two umbrella registry entries dissolved (162 entries)
• The cautious preset says it is hard mode; the title screen says most terms end badly
• Roadmap gates restated in currency the project can actually count
• The sixth ending keeps its mystery
• House typography documents its own rationale where it is enforced

Play at https://criticalwindow.org. Paper kit attached.

v0.3.3: says what it does

A precision release.

• Evidence vocabulary states exactly what each number is: empirically anchored, game-calibrated, or a design constant
• Capability pacing quotes METR's published Time Horizon 1.1 figures verbatim
• Every build ships version.json; deploys verify the live domain serves the deployed commit
• Compute reads against a fixed 2026 baseline
• Public FAQ; session length stated up front; loading state before JavaScript; offline behavior tested against the production build
• Browser suite runs cross-engine and states its scope per test
• Interface icons: Phosphor (MIT), inlined

Play at https://criticalwindow.org. Paper kit attached.

v0.3.2: precision pass

A refinement release on top of the public alpha.

• The evidence pitch now states exactly what it delivers: every number is source-backed or a labeled design choice, and the evidence map opens with five worked examples from source figure to game value
• Values calibrated in tuning wear the design-tuned badge everywhere
• The capability pacing note cites METR's Time Horizon 1.1 update directly
• Release identity (package, citation file, changelog, pinned links) is machine-checked in CI
• Balance invariants are asserted in the test suite instead of published as a snapshot
• One-click recommended first run; honest first-run expectations
• Debrief accepts feedback by email as well as GitHub issues
• Builds display their commit hash next to the data-version hash

Play at https://criticalwindow.org. Paper kit attached.

v0.3.1: the launch-review build

The public alpha, revised after an external zero-context launch review of v0.3.0 found real issues on launch day. All of them are fixed here; the full story is in CHANGELOG.md and docs/BALANCE.md.

Highlights:

• Every public surface now agrees with the source registry, enforced in CI
• SOURCES.md separates external evidence from the design constitution honestly
• docs/EVIDENCE.md: every cited value with its kind and derivation, card premises counted separately
• docs/BALANCE.md: the real ending distributions, reproducible
• Players see source titles, not registry IDs; alignment has translations
• Mobile: ticker wraps, race track behaves, code-split loading
• Content-Security-Policy makes the no-external-calls promise browser-enforced
• Imprint and privacy pages; working contact address; DCO enforced in CI

Play at https://criticalwindow.org. The paper kit PDF is attached.