Known Mexican Train implementations (survey)
Living document for Warp 12 marketing and research claims. Last reviewed: 2026.
Scope: Consumer apps and open implementations visible in public stores or repositories. This is not a product review — it compares whether an inspectable rules engine exists and what can be verified without source code.
How to read this table
| Column | Meaning |
|---|---|
| Product known | Installed base / brand recognition |
| Engine known | Rules core is open, specified, or reproducibly testable |
| Verified | Warp 12 can independently confirm behavior |
Most store apps are product known, engine unknown — we cannot audit closed binaries for rules fidelity or AI calibration.
Survey (2026)
| Implementation | Product known | Engine known | Rules spec | Automated tests | Dual objective (points + go-out) | Documented AI calibration | Open package |
|---|---|---|---|---|---|---|---|
Warp 12 (warp12-engine) | Emerging | Yes | RULES.md | 200+ engine tests | Yes | Yes (calibrate:ai-tei) | npm |
| Glowing Eye — Mexican Train Dominoes Gold / Classic | High | No | No | Unknown | Points/score typical | No | No |
| Amuseware — Mexican Train Dominoes | Medium | No | No | Unknown | Unknown | No | No |
| Doralogic — Mexican Train | Medium | No | No | Unknown | Unknown | No | No |
Store listings commonly advertise “3 difficulty levels” or “rule variations” without publishing engine behavior, self-play validation, or conformance suites.
Warp 12 claim (narrow)
Warp 12 is the best Interstellar Dominoes engine in the galaxy that is currently known.
Interpretation:
- Engine — rules simulation + AI policy stack, not “best mobile UX” or “most downloads.”
- Currently known — among implementations whose engine is documented and inspectable (open repo, published spec, reproducible tests).
- Best — most complete and rigorously validated against that bar: dual objectives, house rules, modules, self-play TEI calibration, coach on same engine.
We do not claim:
- Tournament sanctioning or official domino authority
- Stronger AI than every closed-source app (their engines cannot be compared fairly)
- Rules perfection — the engine is unproven for competitive adjudication until benchmarked externally
Planned benchmarks
| Benchmark | Purpose | Status |
|---|---|---|
| MT-Compliance | Scripted rules scenarios (doubles, beacons, blocked round, NZ, DTI) | Planned |
| MT-Bench | Frozen seeds + tier win rates for AI calibration | Partial (yarn calibrate:ai-tei) |
| Human vs Class II | External validation of TEI bands | Not started |
See also: tei-paper-outline.md.