How it works
DoesItARM is a pipeline: gather signals → resolve a verdict → serve it everywhere → keep it fresh.
1. Gather signals
Section titled “1. Gather signals”Three sources feed each title’s verdict:
- Crowdsourced reports — users submit what happened on their machine (verdict, chip, macOS version, config, date), ProtonDB-style. Cheap, scales on community, captures the long tail.
- Automated test harness (planned) — a Claude-Code-drivable rig installs a title (via
steamcmdor a download), launches it (native, or through Game Porting Toolkit / Whisky / CrossOver), watches for crashes and performance, and records a short clip. Runs on a cloud Apple Silicon fleet for scale. This is the proprietary signal — the “Playwright for games on Mac.” - Known facts — vendor announcements, app binary architecture (
lipo/fileinspection), Homebrew cask metadata, and curated sources. Lowest cost, high confidence where available.
2. Resolve a verdict
Section titled “2. Resolve a verdict”Signals roll up into one current verdict per title — Native / Rosetta 2 / Translation layer / Unsupported — plus a confidence level, the basis (how we know), and the date. Conflicting or stale signals are reconciled with the freshest, highest-confidence source winning. See Compatibility verdicts.
3. Serve it everywhere
Section titled “3. Serve it everywhere”The same verdict is published to three surfaces:
- the free website (humans + AI crawlers),
- the agent API / MCP (planned) — a current, machine-readable answer,
- enterprise readiness reports (planned) — fleet-scale audits against the Rosetta 2 deadline.
4. Keep it fresh
Section titled “4. Keep it fresh”Verdicts decay, so the pipeline re-tests on triggers: a new macOS or Game Porting Toolkit release, an app update, a flood of new crowd reports, or a calendar cadence for popular titles. Freshness is tracked per verdict and is the product’s core promise.
Where the data and the games come from
Section titled “Where the data and the games come from”DoesItARM never licenses or hosts the games. It gets copies to test cheaply — Steam Curator review keys, targeted purchases, and crowdsourced reports from users’ own copies — and publishes only the resulting facts. The legal basis (facts are not copyrightable) and the access tactics are detailed in the Research memos.