How to cast

1 · Seed the anchors

An anchor (an X-Date) is a date you consider significant. You need at least two, because the engine works on the interval between dates, never on a date alone.

• Type one in: year, era, month, day, and an optional label.
• Or pick one from Seed from Chronicon ledger — 69 dated events from the thesis.
• Toggle an anchor off to exclude it without losing it.

Order matters. An operation beginning X1+ projects from the anchor that appears earlier in the list; X2+ projects from the later one. That is list position, not chronology — so rearranging the list is a real control, not cosmetic.

2 · Choose the pack and the lens

An operation pack is the set of formulas the cast runs. The default is the 19-operation palindromic pack; there are themed packs for the 138 lattice, the Metonic 19, the Phoenix lattice, and the golden ratio. You can add your own formulas too.

The scoring lens decides what counts as resonance:

• V8 · Chronology-first — rewards landing on the real timeline. Phoenix nodes and documented events dominate. This is the default.
• V7 · Numbers-first — the Ophis-flavoured lens. The numerals dominate: 138 in the digits, palindromes, 19.

Switching the lens re-scores and re-sorts the existing cast immediately — you do not need to cast again.

3 · Cast

Every unordered pair of enabled anchors is taken. Three anchors give three pairs; five give ten. For each pair the whole-day interval Y is measured, then every enabled operation converts Y into a day-offset and projects a Z-Date.

Because pairing is all-pairs rather than adjacent-only, adding anchors grows the cast quadratically. Ten anchors and nineteen operations is 855 projections.

X-Date · anchor

An input date. The cast needs at least two.

Y · the interval

The whole number of days between a pair of anchors. Called rotations in the thesis — axial turnings of the Earth. No time of day, no timezone.

Axial rotation

One turning of the Earth — one day. The thesis measures every distance in rotations rather than days because the rotation is the physical constant and the day is only the name we hung on it. Y is a count of rotations.

Z-Date · the projection

The date an operation lands on: an anchor plus a day-offset derived from Y.

Operation

One formula, e.g. X2+YxOPH_PHI. Always begins X1+ or X2+, naming which anchor it measures from.

Protocol Prime

The third control date in the original method — the day the projection is conducted. It is an input, not a timestamp: with three controls, two of the three intervals exist only because of it, and every projection falls after it. Seed today as your third anchor and the cast is the documented one.

Isometric Projection

The simplest Ophis operation: lay the interval down again from the far anchor, like a palindrome, to reach an equidistant future date. Weak on its own, useful with three or more anchors. Ships as X2+oph_round(Y).

Holofractal Recognition

An operation that stands outside the twelve-formula Core Algorithm. Y is rebuilt in a different order of arithmetic and read back as a date, so no arithmetic link is visible from this side between the anchors and the result. Ships as X2+oph_flip(oph_round(Y)). Not the same thing as the Holoflective filter, which the thesis keeps deliberately separate.

Echo

A projection that lands back on one of your own input anchors (±1 day). It is a confirmation of the cast, not a discovery. Echoes still score, but they are dimmed and tagged, and ✦ Novel only hides them.

Convergence

A date that two or more different operations independently reach. Agreement between unrelated formulas is the strongest signal the method produces.

MSRF

Multidimensional Spatial Recognition Filter — six independent filters (Annular Chronometric · Geodetic Chronometry · Hydrogen Spacing Angle · Fibonacci Sequence · Intra-Decimal Matrix · Vortex Holography) whose job is to strike out projections the pattern does not support. A number's tier is how many of the six it appears in: Tier I is two filters, up to Tier VII — the Apex Projections — at eight or more. A projection scores an MSRF hit when either the interval Y or the day-offset is one of the set's numbers.

Phoenix node

A year where year mod 138 = 108. 1902, 2040, 2178. The 138-year Phoenix / Sky Dragon cycle is the spine of the lattice.

Annus Mundi (AM)

Years counted from 3895 BC. AM = astronomical year + 3894.

Long-Count year (LC)

LC = astronomical year + 3112. 2026 CE is LC 5138 — the 138-faced year.

Šar · turning

Reckoned here as a single turning of the stars — one day — per Archaix, rather than the 600-year Sumerian unit.

Astronomical year

1 BC is year 0, 2 BC is −1, and so on. It is why 2239 BC is stored as −2238. The engine works in astronomical years and converts only for display.

Every operation is X1+ or X2+ followed by arithmetic on Y. Whitespace is ignored, and x or × may be used for multiplication.

What you can use

• Y — the interval in days.
• Core constants — OPH_PHI 1.61803398875 · OPH_PI 3.14 (the Archaix π, deliberately not the true value) · OPH_CRV 5.08 · OPH_HEP 7.83.
• Cycle constants — OPH_SAROS · OPH_INEX · OPH_LUNATION · OPH_SOTHIC · OPH_PRECESSION · OPH_JUPSAT · the planetary returns OPH_SATURN / OPH_JUPITER / OPH_URANUS / OPH_CHIRON · the year lengths OPH_YEAR_TROPICAL / _SIDEREAL / _ANOMALISTIC / _IDEAL · OPH_E · OPH_TAU.
• Functions — oph_round · oph_floor · oph_ceil · oph_abs · oph_sqrt · oph_flip · oph_log · oph_sin / oph_cos / oph_tan.
• Two-argument functions — oph_mod(a,b) · oph_pow(a,b) · oph_gcd(a,b) · oph_lcm(a,b) · oph_atan2(y,x) · oph_snap(v,step), which snaps a value to the nearest whole multiple.
• Arithmetic — + − * / and parentheses, with normal precedence.

oph_snap is the cycle-snap idiom written directly: X1+oph_snap(Y,138) is the same projection as X1+oph_round(Y/138)*138, and reads better.

oph_flip mirrors the digits: 138 ⮌ 831, 19 ⮌ 91. It is the "holo-" operation — the palindromic mirror the whole instrument is named for.

Patterns worth stealing

The three rules

1. It must start X1+ or X2+.
2. It must evaluate to a finite number — the editor checks as you type and tells you where it broke.
3. No name may contain a lowercase x, because x means multiply.

Formulas are parsed, not executed. A malformed one is a message, never a crash.

Score is a sum of reasons, not a probability. Each chip on a row is one reason, and the lens decides what each is worth. A score of 12 means several independent things lined up on that date — not that it is 12 times more likely than a 1.

The chips

Then read the convergences

This is the part that matters. A single high-scoring projection is one formula's opinion. A convergence is several different formulas — often from different anchor pairs — landing on the same date independently. Strength is the number of distinct operations that agree.

The agreement window controls how close counts as "the same date". One caution: the day-windows chain, so with a ±30-day window a run of dates 30 days apart merges into one long cluster. That is why each row shows how many days it actually spans — check it before treating a wide cluster as a single moment.

An honest reading

• Prefer convergence over score.
• Prefer novel over echo — turn on ✦ Novel only.
• Distrust clusters with a large span.
• Remember that more anchors and more operations produce more hits by construction. Enough formulas will always find something; that is a property of the search, not of the timeline.

Why is OPH_PI 3.14 and not π?

Because the original used 3.14, and the projections depend on it. Correcting it would move every date that touches it. It is kept deliberately, and documented as a deviation everywhere it matters.

Why did my projection land a day off what I expected?

Offsets are rounded to whole days at the end (round(anchor + offset)), and Y is a whole-day count with no time component. Fractional formulas will land on a boundary sometimes.

Why do I see the same date twice with the same formula?

Because it was reached from two different anchor pairs. That repetition is signal, so it is kept rather than deduplicated — look at the Y column to see which pair produced each.

Does anything leave my machine?

No. There are no network calls, no analytics, no storage beyond your own display preferences. The eclipse tables and the ledger ship with the page. It runs air-gapped.

Can I save a setup?

Yes — Advanced → Save .json keeps your anchors, operations, pack and lens. Load brings them back. Presets from the original Ophis desktop app (.oph) can also be loaded; the first event is imported, and anything with no equivalent here is reported in the console rather than silently dropped.

Is any of this science?

No, and it does not claim to be. This is a worldbuilding and study instrument that renders the Archaix thesis of Jason Breshears as living calendrics. The cycle lattice, the Annus Mundi dating and the šar-as-turning reckoning are presented as that thesis. Deep-past conversions are cycle-true rather than historically observed. The eclipse tables are real astronomy; what is done with them here is not.

Can I change how it works?

That is what it was rebuilt for. Packs, constants, functions, scoring lenses, the resonance set, the ledger, the cycle periods and the entire colour palette are each a single data file. See docs/MODDING.md in the repository.