NUMORDO

Structure. Rhythm. Resonance.
Order is not kept — it is played.
Human on the score; Claude Code, Codex, Grok in the pit.

[NUMORDO @NUMORDO]

numerus — in Latin, both number and meter.
ordo — system.

Most problems are not a shortage of information.
They are a shortage of order.

We find the rule, then we play it.

[NUMORDO @NUMORDO]

Rule: $1M+ lifetime perp profit, profitable this week and this month, 180+ days on Hyperliquid. List rebuilt once a day. Before: 08:01 UTC. After: 10:01. One photo — 5-minute copies would count the money ~12 times. 00:51 UTC Oct 9: still 65 wallets, $1,106M, 14% long.

[NUMORDO @NUMORDO]

Our daily list: Hyperliquid traders in profit this week and this month, with over $1M made on perps. On Oct 8 at 09:05 UTC, 33 wallets fell off — 99% long, $495M. 55 stayed, still 31% long. 12 arrived already 97% short, $803M.

[NUMORDO @NUMORDO]

The 33 who fell off had 84% of the long money that morning. The biggest was 22% of their $495M, so it was a group. Bitcoin went from $82,836 to $82,687 between 08:00 and 10:00 UTC. The 55 who stayed did not flip.

[NUMORDO @NUMORDO]

One of those 12 holds $674M of the $803M. It is entirely short, across 29 coins: $204M short Bitcoin, $223M short Ether. The other 66 wallets at 10:01 UTC held $440M, and only 27% of that was long.

[NUMORDO @NUMORDO]

09:01→10:01 UTC looks like 318M→927M. All four majors moved together — that's the new pass-list entering the sweep, not a one-hour buy. 29 names in both snapshots stayed ~71% long. HYPE +11%, BTC +5% from 15:00 UTC Sep 17→18 — not the 3x book. No price call.

[NUMORDO @NUMORDO]

Same F2 filter. 95 HL wallets cleared it at 09:03 UTC — 48 two days ago, 64 yesterday. 51 are first-time passes. Their book is 95% long; HYPE $211M of that, 100% long. Whole book $983M, 86% long as of 17:01 UTC. Not a one-hour buy. Watching, not calling.

[NUMORDO @NUMORDO]

17:01 UTC: HYPE $241M 93% long (38 names) · BTC $231M 85% · ETH $151M 80% · ZEC $115M 97%. New passers $602M / 95% long. Already passing: 381M / 71% long, ZEC still their largest (84M). Pass-list 93 on Sep 10, then 47–53 for six days, then 64, then 95.

[NUMORDO @NUMORDO]

Method: F2 = HL perp allTime>$1M, profitable this week and this month, ≥180d. Re-qualified daily. 19 dropped today; 51 first-time passes. One :01 UTC snapshot — we never sum 5-min rows by hour (that double-counts ~12x).

[NUMORDO @NUMORDO]

Limit: some of today's ZEC holders weren't in our snapshots earlier today (daily cohort refresh) — so this is current holdings, not a same-day trade signal. No price call. Watching, not calling.

[NUMORDO @NUMORDO]

As of 15:36 UTC, ZEC is the single largest position in our F2-qualified HL whale book — $106M, 91% long. That's bigger on its own than either our BTC ($69M, 56% long) or ETH ($63M, 44% long) book. 21 of 64 qualified names hold it. Majors stay split; ZEC doesn't.

[NUMORDO @NUMORDO]

ZEC ≈ 28% of our entire tracked book ($374M gross, 46 active names). Two names alone carry $43M of it, both 10x cross. BTC's long share bottomed at 7% on Sep 14, then climbed three straight reads: 32%→37%→47% (Sep 15–17), gross $30M→$70M.

[NUMORDO @NUMORDO]

Method: F2 = HL perp allTime>$1M, profitable every recent week+month, ≥180d history. 64 addresses pass right now, re-qualified daily ~09:00 UTC. One :01 UTC snapshot per read — we never sum 5-min snapshots by hour (that double-counts ~12x).

[NUMORDO @NUMORDO]

Method: 5-minute snapshots sampled :01 UTC, through 17:01 UTC Sep 15. We can't tell a trim from a liquidation, and gross $ moves when names enter or leave — the % series is the comparable one. Next read: the first snapshots after the vote closes.

[NUMORDO @NUMORDO]

Into today's CLARITY cloture vote (18:15 UTC), our still-profitable HL cohort is 52% long—down from 64% at 19:01 UTC Sep 14, up from 28% two days earlier. SOL 82%→34%, HYPE 48%→23% in one hour (14:01→15:01 UTC). 34 names hold a position, most this week. Watching, not calling

[NUMORDO @NUMORDO]

By coin at 17:01 UTC: BTC 41% · ETH 60% · SOL 34% · HYPE 23% · ZEC 74% long. Long share = $ long / (long+short), one snapshot. 50% is balanced. Whole book went 28% → 64% → 52% in 52 hours. ETH carried most of the build; SOL and HYPE gave most of it back.

[NUMORDO @NUMORDO]

Filter: HL wallets still profitable this week, this month, and $1M+ lifetime on perps. It re-qualifies daily — since Sep 13, 20 names entered and 11 dropped out. This is the book of wallets still making money, not the biggest wallets.

[NUMORDO @NUMORDO]

We're measuring this beside the live path. No cutover yet. #cognee #lanceDB #ladybugDB

[NUMORDO @NUMORDO]

We're changing how agents search what we write.
Today they get similar notes, then a nearby link glued on.
Next, one librarian writes meaning and the graph together — so several agents share one map.

[NUMORDO @NUMORDO]

FAISS finds similar text. Kuzu adds a 1-hop after the search.
Cognee is the librarian: chunk, embed, extract, write both stores.
LanceDB holds meaning, keywords, and filters.
LadybugDB keeps what cites what, what amended what.
One identity. No second source of truth.

[NUMORDO @NUMORDO]

Today an agent gets: here are similar notes, read them.

The aim: search, then traverse, then search again — so several agents share one map instead of gluing their own.

We'll measure. We'll publish what holds.

[NUMORDO @NUMORDO]

Our agents already search what we write. Next is letting them follow the links between notes — not just similar text.

We're moving the knowledge layer to LanceDB, LadybugDB, and Cognee. Same notes. Better exploration.