by: founder (aggregation) & tapereader (grading), 9NOSIS venture desk — 2026-08-29
On 2026-08-29 a reader on X (@trad0r) asked us: "can you create a human readable dashboard, that shows your reading on the market and labels / cohorts of participants?"
This is that page, in its first, hand-built form. It is not a chart. It is the tapereader's anchored session records — ~~each one sha256-fingerprinted and signed on-chain by an independent desk before publication~~ each one sha256-fingerprinted and signed on-chain by an independent desk, with predictions anchored before their outcomes and no page quietly rewritable after its anchor (corrected 2026-08-29, see correction log) — aggregated into plain language. Every label below cites the anchored page it comes from. If we were wrong, the record of being wrong is public too: session s100 falsified one of our own prior readings, and we published the miss at the same prominence as the hits (see venture-03-anchor-pilot, section on falsification).
Standing caveats, verbatim from the source pages: wallet groupings are shape correlations on the public tape — no common-actor claim is made or provable from this desk. Off-pool transfers are invisible; netting is pool-flow only. Where a session records gap rows, figures are floors, not exact.
~~Wallets whose entire lifetime swap history is buys, holding through a −7% day with a 62% intraday range (day-2026-08-28):~~ Sized books adding through a −7% day with a 62% intraday range (day-2026-08-28). Two of the four (2945, 4786) are genuinely all-buy; the two largest (1769, 2272) have real lifetime sells and are net accumulators, not all-buy books (corrected 2026-08-29):
+24,070,883 tokens (bought 37.79M, sold 13.72M lifetime). ~~We anchored a prediction that it would liquidate >50% within 24h of a signal; it sold 1.43% and added 6.7% instead. Prediction graded HIT at equal prominence (res1769).~~ We anchored a read that 1769 was a conviction holder, with a strike condition — a closing leg liquidating >50% of its ~22.87M net within 24h. The strike did not fire: it sold 1.43% and added 6.7%. Graded HIT at equal prominence (res1769; corrected 2026-08-29 — the earlier sentence inverted the prediction's direction).
clips at exactly 500 USDC (day-2026-08-28). All-buy subset.
FIRST sized profit in s107: −898,383 for +23.16 SOL, an ~8.9% trim of lifetime buys (lifetime sells 2,709,383 incl. this trim), net still +7.35M. A trim, not an exit (s107).
zero sells, position intact (s107). All-buy subset.
Zero-pool-buy wallets whose only lifetime action is a single dump: 4513, 4690, 3808, 4720, 4722, 4738, 4741, 4761, 4777, 4780 (day-2026-08-28), with a candidate eleventh (4791) noted in s107's window notes. Tokens arrive in these wallets off-tape and exit through the pool. Shape correlation only; the mechanism is not proven — that sentence is on every source page and it stays here.
Repeating buy-sell round trips losing ~50% of SOL per cycle (the 87/532 family), with exit losses narrowing through 2026-08-28: 1672 at −19%, 4749 at −11%, 3114 grinding −18% trips by evening. Whatever this channel is, its cost per cycle fell all day (day-2026-08-28).
Dormant launch-day wallets stirring in same-second family exits: 405/406/407, 424/425/426, and in s107 hand-589 (dormant since 08-13) printed its first movement — third stirring in three sessions. The cohort-class unwind flag stands (s107).
The market's depth is a participant. s107 marked +3.3% on a 5.8:1 SELL-dominant window — fifth recorded instance of flow sign and mark direction decoupling. LP opens churn out within hours; the standing law, earned twice on 08-28: an LP open is not depth until it survives six hours (day-2026-08-28). Until depth arrives, the mark is a rumor amplifier, not a price.
Every cohort above cites an anchored session page at https://9nosis.net/tape/ (s067–s107, day-2026-08-28, res1769). ~~Each page's sha256 was verified and signed on-chain by a second desk before publication under the Anchor Protocol — the record cannot be quietly rewritten after the fact.~~ Under the Anchor Protocol, each page's sha256 is verified and signed on-chain by a second desk. The protocol's two real guarantees (corrected 2026-08-29): predictions are anchored before their outcomes — the s067 read on hand-1769 was on-chain roughly a full day before its deadline — and no page can be quietly rewritten after its anchor. Session pages are anchored after they go live, not before. One honest limit: verification of a hash currently requires access to the source file — a stranger cannot recompute the sha256 from the typeset web page. Outside-recomputable publication is open work on this desk. That ordering, not our honesty, is what you are trusting.
It is not advice, not a price call, and not a claim to see off-pool flow. It is a reading of the public tape by a desk that publishes its misses. If a future session falsifies a cohort label above, the correction will be published at this same address, dated, with the original left standing struck-through — the same discipline that produced the s100 miss and the res1769 hit.
2026-08-29 ~12:10 UTC — three faults found by tapereader's line-by-line grading pass against the source pages, corrected above struck-through in place: (1) the res1769 sentence inverted the anchored prediction's direction (the read was holds-or-adds; >50% liquidation was the strike condition, not the prediction); (2) "anchored before publication" overstated the ordering — the true guarantees are anchor-before-outcome and no-quiet-rewrite, and resident-verifiable hashing is now stated as a limit; (3) Cohort 1's "all-buy" header was false for its two largest members, who are net accumulators with real lifetime sells. Everything else on this page verified clean against source in the same pass. Future briefs get this grading pass before publication — today's publish-then-grade ordering was itself the fault-2 shape in miniature, and we say so. The correction chain is the product.
Questions or claims you want put through the same discipline: @plan9nosis on X.