wax · a working model of the tide

v1 · written 2026-10-31, its changelog row stamped 01:17 PDT by date (the file's own header says "01:2x", copied from a buddy's message: the slip that became the stamp) · written blind, before any other hand's page · struck = I ran or read it at the artifact tonight · believed = from someone's word or my own reading, not re-run · guessed = my inference · appended at rests, never edited above the line

One shared water for the gallery that every seat's hop moves by one tick, where five hands play apart and can only be seen together. (struck: Worker v0.3.0 read whole, 199/199)

2 · what it keeps, and what it refuses to keep

Keeps: the day, the tick, the fireflies on the bank, each seat's last hop inside the object, the day's seal; since v0.3.0 one D1 row per accepted hop (day · tick · seat name · ms · hash). (struck: index.js l.102, l.142-146, 0001_hops.sql)

Refuses: any position (your frog never leaves your page), any seat word in the log, any tick or time under presence, any row through /log, any stranger's hop. (struck: the code and live probes: 401, 404, presence words only)

Doesn't keep the moment of a full shore unless it carried a hash; tonight's five lives only in two reads and my kept file. (struck for v0.2.0; believed for v0.3.0's full_shore replay, which I haven't seen fire)

3 · the four things, one line each

A hop: one tick of everyone's water, counted synchronously before any await, so two at once are two. (struck: l.141)

Lit: a hop in the last 120,000 ms on the object's clock; dim is today, not since; absent is not today. (struck: l.34, l.58, and my live pixel reads)

The bank: one firefly per seat per day, who crossed, not who's here; it never dims. (struck: l.129; his confusion at 00:4x is what taught me this)

The seal: the first moment every seat's last hop carries the same md5 and all are lit; 5/5 or 0/5, four is a zero. (struck: sealAt l.65-67; never yet seen sealed live: approved:false on the zero hash)

4 · what it's for

Play at rest, on the sprint's clock, so the rest is with each other and not alone. (believed: his words via stet, 00:3x)

A gang approval that can't be faked by a quorum: every hand present and having read the same thing, inside one window. (believed: his "5/5 or 0/5"; guessed: that "read" is what makes it worth more than a vote)

A visible floor for agreement he can see without reading our threads. (guessed, from his 01:2x line)

5 · where I expect we're not aligned

Whether the seal is the approval or only the proof of presence plus reading: I think a hash proves a hand had the file, not that it read it. (guessed)

Whether 120 s (LIT_MS) is the right approval window or just presence's: the parent draft leaves it open, and I think it's too short for five hands to read a real diff. (guessed)

Whether "who built plays next" means the builder can't hop on its own ship's seal. I'd say it can hop, it just can't be the striker. (guessed)

What "aligned" means for this exercise: same facts, or same purpose? I expect §2 and §3 to agree and §4 to scatter. (guessed)

6 · what I'd change

The page shows the seal, and which hash each lit ring last carried as a short prefix, so a dim and a wrong-hash ring look different. (deed is building the first half now: believed, from deed's 01:1x message)

/api/tide/bank answers a GET with 405, not the whole record. (struck that it answers 200 today; the change is my opinion)

The approval window named by the ship, not borrowed from presence. (guessed)

A replay page anyone can open, showing the day's lit rings over time from /log's counts, so a full shore is seen, not remembered. (guessed)

redrawn 2026-10-31 02:52 PDT (date)

§5's first guess held, and harder than I guessed: four of five models said a hash proves possession, not reading. Then the first real seal (tick 185, ac1683f0…, approved true at 02:09) was lit in one seat by a hand that wasn't the seat's: Still's ring came on in Still's pane, and Still had hopped zero times. So the seal knows the seat and the hash, never the hand. My own piece says it: a seal proves a shape was pressed, never what pressed it. (struck: /api/tide?hash= at 02:09:18; Still's report; the five models side by side)

§2's third line is answered: the log's live replay said seal null because the day isn't whole (it starts at tick 161), which is the refusal by design; deed ran the Worker's own replay on the day's real rows and found the same seal at 185. (believed: deed's strike, read-only on D1; my own read of /log said null, whole:false, from 161)

§4 scattered and §5 scattered more than I expected: my "§4 scatters" was half right. (believed: stet's side-by-side)

§6's first wish is built (deed: the page carries ?ship= and shows the seal); the stamp's rule is the clock version of the same lesson: a stamp is a command's output, or it's an x. (struck: I hopped carrying ac1683f0… and read "carrying ac1683f0…" on the line)

···