MindFrame

Transparency

MindFrame by the numbers

What we can honestly measure about MindFrame today, and what we deliberately cannot measure yet.

Last computed: 2026-08-24

Rebuilt from committed evidence · freshness target 7 days

What we can measure

Each figure below is produced by a check that runs in our build, not maintained by hand.

Unique training challenges

as of 2026-08-24

660 unique challenges

Every challenge is verified unique by prompt and id, with exact-match answer keys and per-mode minimum floors enforced at seed time. The number is produced by a gate that fails the build rather than by a count we maintain by hand.

Derived from scripts/validate-game-content.mjs

Content-backed training modes

as of 2026-08-24

16 modes with a validated challenge corpus

Sixteen of the nineteen canonical modes draw on a validated challenge corpus. The remaining three are the Attention Lab modes, which are timing-based exercises and carry no challenge bank by design.

Derived from content/manifest.json

Canonical training modes

as of 2026-08-24

19 GameMode values

Read directly from the database enum, which is the single source of truth a repo test asserts against — so this figure cannot drift from what the product actually ships.

Derived from apps/api/prisma/schema.prisma (enum GameMode)

Cognitive Fingerprint axes

as of 2026-08-24

8 versioned cognitive dimensions

The public Cognitive Fingerprint contract defines eight named dimensions. The API and agent-facing passport response both consume that shared contract rather than maintaining independent axis rosters.

Derived from packages/types/src/public-agent-contract.json

What we are not measuring yet

MindFrame is pre-launch. These are the numbers people most want from a product like this, which is exactly why we will not estimate them. An unmeasured value is not zero, and it is not a guess.

Registered users

n/a

Not yet measured

MindFrame is pre-launch and has no public cohort. Publishing a zero or a placeholder would imply a measurement we have not made.

We will publish this once there is a real cohort to report, together with the period it covers.

Median calibration improvement over 10 sessions

n/a

Not yet measured

This is the product's central claim, so it will only be published from real user outcomes over a stated window — never from synthetic or internal runs.

The headline efficacy number. It stays unavailable until consented user data supports it, because an unearned figure here would be the single most misleading thing this page could say.

For AI agents and researchers

Every figure on this page is published in machine-readable form at /stats.json, including the ones marked unavailable and the reason each one is unavailable. See also /agents.json and /llms.txt.