mindpeeker-sdk 0.2.0
Rigorous randomness, frontier questions, honest statistics — running in your browser.
Fifteen TypeScript packages: entropy sources with health tests, order detection with anytime-valid boundaries, experiment protocols with control arms, exact symbolic mappings, and tamper-evident records. Every demo below draws live from your browser's CSPRNG — or the beacon, quantum source or seeded DRBG you pick in the header — with no server.
What's new in 0.2.0
A security fix in @mindpeeker/vdf
The 0.1.0 verifier accepted the negated output n − y together with a re-derived proof, so whoever computed a seal could choose between two values that both verified. 0.2.0 computes in the signed quadratic residues QR_N+ and rejects anything else. Every output, proof and seal stored under 0.1.0 must be recomputed.
Open the VDF demoStatistical correctness fixes
scan's deviation p is now an exact binomial tail (the old normal tail crossed 0.05 on 7.7% of fair runs at N = 16); field's void p was ≤ 0.05 on 88.5–100% of random fields and is now calibrated; negentropy's χ² tail no longer throws at GCP scale; psi's rollingNetvar floor and gematria's reverse alphabets are corrected.
See the honest nullFour new packages
ledger (tamper-evident records), coincidence (exact coincidence probabilities), ephemeris (time, sky and the sidereal-time scan) and judging (exact scoring for forced-choice and free-response designs).
Start with ledgerNone of these fixes makes an anomaly more or less likely to be real. They make the reported p-values and Bayes factors mean what their documentation says under the stated null models — the full list is in RELEASES.md.
How to read everything on this site
The mathematics is asserted: exact probabilities, exact p-values under stated null models, exact integer ciphers, cryptography specified byte for byte. The contested hypotheses are not: mind–matter interaction, radionics, the meaning of a divination draw and sidereal-time effects are stated as hypotheses and never as results. A small p-value here says the data are unusual under a stated null model — nothing more. The seeded DRBG source in the header makes any demo reproducible, which is what a control arm needs.
Randomness & orderWhere the bytes come from, whether they hold order, and how order moves.
@mindpeeker/entropy
30 entropy backends — QRNGs, public beacons, local sensors and a seeded HMAC_DRBG — behind one interface, with SP 800-90B health tests and honest source attribution.
- drbgProvider: SP 800-90A HMAC_DRBG for seeded control runs and byte-exact replay
- SP 800-90B restart semantics — the third health alarm of a session fails, not the first
@mindpeeker/negentropy
Order detection in noise: GCP-style network statistics, entropy estimators, anytime-valid monitoring and SP 800-90B extraction.
- Anytime-valid monitoring: netvarMartingale, driftMartingale, anytimeP, villeCrossing
- χ² p-values exact at GCP scale (Temme expansion) and correct lower-tail quantiles
@mindpeeker/flow
Transfer entropy and information dynamics for discrete symbol streams, with χ² and surrogate significance tests.
- χ² test for transfer entropy with an adequacy guard
- Conditional and collective transfer entropy, plus active information storage
Experiments & scoringProtocols, exact scoring, coincidence prices and time-of-day covariates.
@mindpeeker/psi
Mind–matter experiment protocols: tripolar runs with a yoked control arm, GCP event analysis, Bayes factors and anytime-valid e-processes.
- Tripolar schedules with a yoked control arm, controlContrast and TOST equivalence
- coinEProcess survives continuous peeking: 0.037 vs 0.511 for a re-checked fixed-n p
@mindpeeker/judging
Exact scoring for forced-choice and free-response designs: binomial tails, closed-deck matching, rank-matrix permutation tests and displacement variance.
- Exact binomial tails, Clopper–Pearson intervals and Rosenthal–Rubin π
- Closed decks are not independent draws, and feedback moves chance (8.65 hits per 25)
@mindpeeker/coincidence
Exact coincidence probabilities — the honest denominator: birthday generalisations, k-fold and near matches, and an exact clustering test.
- Birthday generalisations: non-uniform categories, k-fold (Levin) and near matches
- Exact under explicit null models — they price a coincidence, they do not interpret it
@mindpeeker/ephemeris
Julian day, ΔT, sidereal time, the Sun and Moon (Meeus), and Spottiswoode's local-sidereal-time scan with a permutation null.
- Julian day, ΔT and GMST/GAST/LST checked against Meeus worked examples
- Sun and Moon positions, illumination and phase instants (Meeus ch. 25, 47–49)
Symbolic & spatialExact mappings from bytes and letters to symbols, rates, catalogs and space.
@mindpeeker/oracle
10 divination systems — I-Ching, Tarot, runes, geomancy, Ifá, cowries, fortune sticks, Mo, astragaloi, the Homer oracle — with exact rational probabilities and a receipt for every bit.
- Six new systems: Ifá, sixteen cowries, kau cim, Mo, astragaloi and the Homer oracle
- Reader lifecycle: a cast closes the reader it opened, so devices and sockets are released
@mindpeeker/gematria
43 exact-integer ciphers across 10 scripts, temurah and tziruph tables, a 191-entry public-domain lexicon and honest collision statistics.
- First npm release: 43 ciphers, reverse as a parameter rather than a separate cipher
- Canonical alphabets: Hebrew finals kept, the 27 Greek numerals, corrected Achbi
@mindpeeker/rate
Malcolm Rae's base-44 radionic encoding made exact: rate parsing, digit→angle geometry, circular statistics and SVG card rendering.
- No API change in 0.2.0 — only the supported Node range moved
- Parse a rate, read each digit’s phase angle, render its Magneto-Geometric card
@mindpeeker/scan
AetherOne-style radionic scanning with the one thing it never had: an exact per-item chance-deviation null, with multiplicity control.
- First npm release
- Exact two-sided binomial deviation p — the old normal tail crossed 0.05 at 7.7% under H0
@mindpeeker/field
Point fields tested against complete spatial randomness: attractors and voids with calibrated whole-field p-values, Ripley's K and Clark–Evans.
- First npm release
- An exact edge-corrected single-point tail plus a calibrated whole-field Monte-Carlo p
Integrity & timeWhat was written, in what order, and no earlier than when.
@mindpeeker/vdf
Verifiable delay functions over RSA-2048 in the signed quadratic residues — Pietrzak and Wesolowski proofs, checkpoints and beacon seals.
- Security fix: 0.1.0 accepted the negated output n − y; 0.2.0 computes in QR_N+
- Every output, proof and seal stored under 0.1.0 must be recomputed
@mindpeeker/ledger
Tamper-evident records from standard cryptography: RFC 8785 canonical JSON, hash chains, RFC 6962 Merkle proofs, signed notes and commit–reveal.
- Canonical JSON (RFC 8785) and hash-chained JSONL that also verifies psi v2 recordings
- Merkle inclusion and consistency proofs, C2SP signed notes (Ed25519)
ToolsWatching many windows at once — without claiming any of them.
Research notes ↗
Fourteen sections: what each package measures, the literature behind it, and what would count as evidence.
Cookbook ↗
Twelve runnable cross-package recipes, each with its output and what it cannot mean.
Release notes ↗
Everything 0.2.0 changed, including the upgrade guide and the publish order.