Skip to content
STP-KASPublic

About

Map of every Kaspa TN10 storm (testnet only). Thanks to Kaspa Pulse (@gokugalax).

Resources

Stars

0 stars

Watchers

0 watching

Forks

Repository files navigation

Experimental. We are just trying this. Disclaimer · Kaspa Testnet-10 only; tKAS has no value.

On 15 Nov 2026 we call it a day until further notice. The ordered list of every row is RECAP.md. This page does not arm a run.

tn-master: every TN10 storm, one map

TN10 storms are load tests we ran on Kaspa's Testnet-10 from 18 Sep 2026 on: many senders pushing simple transactions at public and private nodes, then covenant families, then counting what the chain actually accepted. This repo holds no raw data. It links into each storm repo.

Read this repo in this order

  1. DISCLAIMER.md.
  2. RECAP.md: every index row, in index order, one outcome line. Not a new measurement.
  3. repos/INDEX.md: class, status, lesson, and finding numbers.
  4. FINDINGS.md: items 1 to 109, no missing number.
  5. The headline table on this page: the detailed numbers. It is not in clock order through 14 Nov.
  6. history/TIMELINE.md: one line per day through pre-run 23, then one line for 14–15 Nov.
  7. OBSERVED-FOR-DEVELOPERS.md: notes from pre-run 9 on. Observations only. Our methods may be wrong.

Review of 34 pushbacks (O1–O34), re-read 15 Nov 2026: pushback-review/.

Every neuron uploads. A neuron is one storm repo, one clock, one question. The day it is created, it gets a row here. The hour the clock closes, the result or the miss is written here. Success and failure both stay. The page is docs/NEURON.md. Read it before a test starts.

Thank you, Kaspa Pulse (@gokugalax, kaspapulse.com)

Since 4 Oct 2026 Kaspa Pulse has counted our storms from the outside, on his own two nodes, and he shaped how we run and count them. These pages rest on his work:

  • The node-dead rule. A sender stops when its node is dead (RPC failing, not synced, lag over 60 s, or DAA still for 30 s); a failed submit is a reject, never "sent". He spotted our senders firing at a dead node on 9-10 Oct.
  • The chain count right after END, on non-archival nodes, before pruning removes the blocks. Every storm since 10 Oct has a COUNT step because of him.
  • The "accepted in blocks" counter: independent counts by Kaspa Pulse — accepted in block counts on two external nodes, per UTC minute, by block time. It is the cross-check for every headline here (METRICS.md).
  • The base-load figure: TN10 carries about 56 tx/s of base load. Every network count on these pages includes it.

Every input he gave and what we changed: credits/KASPA-PULSE.md.

Headline per storm

Best number per storm, with its counter. Pre-run windows run from the repo's own logs/PRE-RUN-OPEN.md to the runner's end (or FAILED); history/TIMELINE.md gives longer Brussels windows that include setup. Counters differ between storms; read METRICS.md before comparing rows. Every network count includes TN10 base load of about 56 tx/s (Kaspa Pulse). This table is not in clock order through 14 Nov. RECAP.md is the ordered list.

Storm Best number (reported) Counter Repo
18-21 Sep, farm mining not a tx rate — tn10-storm-2026-09-18-mining-public
25 Sep, round 1 break test 6,466 tx/s mean, 10-s peak 9,274 Processed (overstates) tn10-storm-2026-09-25-r1
25 Sep, Build wallet load 1,162 tx/s offered for 0.29 s (340 transactions), all included; no local node Own wallet, offered and RPC-accepted; inclusion checked tn10-storm-2026-09-25-build
25-26 Sep, rounds 2-6 1,219 tx/s for 1 h 32 min (round 4) Accepted (own runner) tn10-storm-2026-09-25
26 Sep, round 7 267.2/s unique vs 357.2/s Processed Both tn10-storm-2026-09-26-r7
26 Sep, round 8 (16:34–18:00 CEST under storm) 1,191 of 1,196 covenant instances completed under storm; 7,894 of 7,894 illegal spends rejected; network 1,120.8 tx/s selected-chain accepted in the full-rate 26 min Own covenant runner; network selected-chain accepted tn10-storm-2026-09-26-r8
27 Sep, desk CPU miners not a tx rate (no headline number recorded) — tn10-storm-2026-09-27-desk-miners-public
1-3 Oct, box storm best 1 min 4,253 tx/s; 58,905,910 included Included (own) tn10-storm-2026-10-01
2 Oct, full gusto about 112 tx/s (one snapshot) Accepted (sender) tn10-storm-2026-10-02-public
6-7 Oct, Build desk 2,207 tx/s seen accepted for 6 h 1 min Seen accepted tn10-storm-2026-10-06
7 Oct, mass ceiling about 3,040-3,080 payments/s at 10 bps Block reading (capacity) tn10-storm-2026-10-07 (run); note: what-limits-tx-rate
8 Oct, checkout 2,191 submit / 2,143 accept tx/s Submit / local accept tn10-storm-2026-10-08
9 Oct day, keel max 2,435, avg 1,627 tx/s unique accepted Network unique tn10-storm-2026-10-09
9-10 Oct night 7,541,103 in blocks; peak minute 1,223 tx/s Kaspa Pulse, accepted in blocks tn10-storm-2026-10-09
10 Oct, public-node test (12:23–14:18 UTC) about 3,168 tx/s submitted (about 2,969 accepted) at peak on 22 senders; clean hold about 880–985 tx/s on 14 senders Bot submitted / accepted (own senders, public nodes) tn10-storm-2026-10-10-public-nodes
10-11 Oct (Sat 18:10 - Sun 02:35 UTC) MAX step median 2,642.4 tx/s (n=9; corrected 11 Oct, was 2,644.9); Kaspa Pulse: 32.3M accepted, best minute 2,785 (21:20Z), 18:10-02:35Z Network unique (headline) tn10-storm-2026-10-10
11 Oct (Sun 07:30-12:33 UTC) peak complete minute 3,583.1 (11:28Z); mean 2,833.1; median 3,107.1 (07:30-12:33Z); Kaspa Pulse: mean 2,844, median 3,055, best minute 3,646 (11:28Z), 51.2M accepted (07:30-12:30Z) Network unique (headline); Kaspa Pulse method live as cross-check tn10-storm-2026-10-11
29 Oct (Thu 19:00-20:00 UTC) total hops/s: 1 family 0.81, 2 families 1.96, 4 families 3.24; single chain under 1 hop/s (20 Oct baseline ~0.7) Hops/s, covenant families; Kaspa Pulse independent count tn10-storm-2026-10-29 (review)
7 Nov (Sat 19:00-21:00 UTC, failed) window did not run. No genesis, no sends, no count. Pre-tests still stand and are not that window Hops/s per covenant family. Desk sends, mini counts tn10-storm-2026-11-07. Record: tn10-storm-2026-11-08
8 Nov rehearsal (Sun 00:40–04:50 UTC) 32-family mean 2,922.8 hops/family over 2,700 s, under the 3,078 line; sixteen-family mean 1,111.8 over 900 s, every family above 1,026 Desk accepted hops. One reading. Not Pulse's count tn10-storm-2026-11-08
8 Nov confirmation (Sun 12:00–16:10 UTC) 32-family mean 3,006.8 over 2,700 s, under the 3,078 line; sixteen mean 1,111.1 over 900 s, every family above 1,026. Both publishers fired Desk and Mini, not averaged. Not Pulse's count tn10-storm-2026-11-08-confirm
11–12 Nov, Wednesday pre-runs 1–9 In our later runs, when we pushed over capacity, our desk node accepted about 110,000–132,000 tx/min (pre-run 8). Submit-OK is not inclusion. Pulse has not counted it Desk submit-OK; desk-node inclusion in pre-runs 8–9 tn10-storm-2026-11-11-prerun-2 (index)
12 Nov, pre-run 10 (07:24–07:53 UTC) Payments included as sent at 1,013, 1,500 and 1,800 tx/s; 1,607,515 payment submit-OKs, all included by the end of the tail. Cheap chained tiers already slow at 1,013 Desk submit-OK; desk-node inclusion. Not Pulse's count tn10-storm-2026-11-11-prerun-10
12 Nov, pre-run 11 (08:59–09:46 UTC) Payments included as sent at 400, 700 and 1,013 tx/s; 923,812 payment submit-OKs, all included. In our read the floor tier followed the ready mass in our node's mempool more than our offer Desk submit-OK; desk-node inclusion; read-only mempool probe. Not Pulse's count tn10-storm-2026-11-11-prerun-11
12 Nov, pre-run 12 (10:37–11:46 UTC) No payment load. Every OK family hop included (40,440); about 49 hops per family a minute with the outside heavy stream absent, 76–92 percent of that (per-family medians) with it present Desk-node hops per family a minute, percent of that rate. Not Pulse's count tn10-storm-2026-11-11-prerun-12
12 Nov, pre-run 13 (13:10–15:17 UTC, failed before RETURN) Our own heavy lane against five tiers: all 40,830 lane transactions and every OK hop included; tiers 95–101 percent of BASE at N = 10, 8–67 at N ≥ 40. Failed on a coin-count abort at 15:17:53Z Desk-node landings, percent of BASE. Not Pulse's count tn10-storm-2026-11-11-prerun-13
12 Nov, pre-run 14 (18:19–19:57 UTC) A hop at twice our lane's rate (the 2.00 row) kept 96–99 percent of BASE (76–88 at the lane's rate); all 13,150 lane transactions included Desk-node hops, percent of BASE. Not Pulse's count tn10-storm-2026-11-11-prerun-14
12 Nov, pre-run 15 (20:44–22:10 UTC) About twice our uncapped lane's rate kept 95–98 percent of BASE, about three times needed at the heaviest step; all 15,831 lane transactions included Desk-node hops, percent of BASE. Not Pulse's count tn10-storm-2026-11-11-prerun-15
12–13 Nov, pre-run 16 (12 Nov 23:10 – 13 Nov 00:37 UTC) Fee duel, both sides ours: settled near twice the lane's rate (median ratio 1.96), no spiral; all 22,258 lane transactions included Duel log (sompi/gram); desk-node landings. Not Pulse's count tn10-storm-2026-11-11-prerun-16
13 Nov, pre-run 17 (01:28–02:12 UTC, failed) Two equal heavy lanes of ours on one rule stepped down 150 → 120 → 100 in four minutes. Failed three minutes into HOLD-3 on the family key's abort line at 02:12:41Z (03:12 Brussels; refilled too little, our error) Duel log; desk-node landings. Not Pulse's count tn10-storm-2026-11-11-prerun-17
13 Nov, pre-run 18 (03:06–04:23 UTC) Greedy duel: 150 → 1,405 sompi/gram in about 21 minutes while both lanes together landed about the same (median 548 a minute at the reference, 562 and 560 in the greedy parts) Duel log; desk-node lane landings a minute. Not Pulse's count tn10-storm-2026-11-11-prerun-18
13 Nov, pre-run 19 (05:08–05:44 UTC) Greedy alone: 150 → 719 in 7 minutes with flat landings (median 518 a minute against a reference of 513): a ceiling, not competition Duel log; desk-node lane landings a minute. Not Pulse's count tn10-storm-2026-11-11-prerun-19
13 Nov, pre-run 20 (07:53–08:43 UTC) A never-down rule ratcheted 150 → 368; at a fixed 460 the template stayed full with 8 chains (about 472,000 mass) and shrank with 30 (about 348,000) Duel log; desk-node template mass (read every 3 s). Not Pulse's count tn10-storm-2026-11-11-prerun-20
13 Nov, pre-run 21 (10:21–10:35 UTC, void) Void, no usable data: the desk and the Mini lost internet from about 10:25Z to 10:33Z (a robot vacuum knocked the network switch). No findings. Spend still recorded: 207 lane transactions accepted (0 rejected, 0 lost), about 540 tKAS in all Spend only, from its LESSONS (no network reading) tn10-storm-2026-11-11-prerun-21
13 Nov, pre-run 22 (11:35–12:10 UTC) Cap path ran live as planned; our lane landed 235–239 a minute at 460 and 201–220 at 2,000 (8 chains); templates under 400,000 mass in 7 / 20 / 38 percent of samples with 8 / 15 / 30 chains Desk-node lane landings a minute; desk-node template. Not Pulse's count tn10-storm-2026-11-11-prerun-22
13 Nov, pre-run 23 (16:08–16:57 UTC) Templates under 400,000 mass in 7 / 30 / 88 percent of samples with 8 / 15 / 30 of ours ready (likely tied to the ready count, not proven); 233–244 landed a minute at 2,000 against 231–240 at 460 Desk-node template; desk-node lane landings a minute. Not Pulse's count tn10-storm-2026-11-11-prerun-23
14 Nov, pre-run 28, retry 1 (07:55–08:12 UTC) Soft partition on our two nodes: a 3-minute desk–Mini peer cut and restore left both pools empty at the end of the 5-minute window, our three transactions landed before, during and after (met, retry 1). Before it, Mini's pool held 2,705 old entries, at least 10 already accepted; a clean restart of Mini cleared them all Each node's own pool and UTXO index, sampled every 10 s; our transactions' outputs in the desk UTXO index tn10-storm-2026-11-11-prerun-28
14 Nov, pre-run 29 (03:48–04:01 UTC) Public comparison on a quiet network: our desk and a verified public node (3 distinct public ids of 9 names) read the same fee estimate (100 / 100 / 100) and the same small tip blocks in all six samples; both normal transactions landed (met, retry 1) Each node's own fee estimate, mempool size and tip block, read in the same minute; our transactions' outputs in the desk UTXO set tn10-storm-2026-11-11-prerun-29
14 Nov, pre-run 30, IBD part (04:03–06:05 UTC) A fresh v2.1.0 node, disposable, on our desk synced TN10 from nothing in 1 h 59 min 44 s: headers 12 min (1.3 million), UTXO set 49 min to receive (337.8 million entries) and 25 min to import, blocks 27 min; same DAA and pruning point as our main node after; directory 46 GB, deleted (met, retry 1) The disposable node's own log and isSynced, compared with the main desk node tn10-storm-2026-11-11-prerun-30
14 Nov, pre-run 30, no-disk part (03:11–03:15 UTC) v2.1.0 edges on our desk node: a coinbase-shaped transaction was refused at the mempool; a 3-input spend of our own script was accepted and an 840-input one was rejected at the transient mass limit (1,000,000), which binds before the 500,000 compute limit for big signature scripts; ZK verify not met (no artifact); IBD not run (met for coinbase and the script-mass pair, retry 1) The desk node's own accept or error string for each single submit tn10-storm-2026-11-11-prerun-30
14 Nov, pre-run 26 (02:25–02:56 UTC) Held 2× fee step against our own heavy lane: floor hops a minute 649.7 at baseline, 611.7 with the lane on, 623.3 with our mix at 2× the lane, 631.3 at zero; 60 of 60 mints and 16 of 16 guest carriers; the 2× step halved the lane's share of the template; lane stopped at its zero (met, retry 0) Desk-node accepted hops and mints (coordinator read; Mini read every minute), the desk template sampler and fee estimate, guest carrier ids among the desk node's accepted ids. Not Pulse's count tn10-storm-2026-11-11-prerun-26
14 Nov, pre-run 25 (01:02–01:34 UTC) Speed step, family count 4 → 8: a covenant hop, a KCC-20 mint and a tic-tac-toe guest carrier still all landed in one window: 14,339 hops, 30 of 30 mints and 8 of 8 carriers of one match accepted (met, retry 0); no component fell behind Desk-node accepted hops and mints (coordinator read; Mini read the same per minute) and guest carrier ids found among the desk node's accepted ids. Not Pulse's count tn10-storm-2026-11-11-prerun-25
14 Nov, pre-run 24 retry 1 (23:46–00:17 UTC) A covenant hop, a KCC-20 mint and a tic-tac-toe guest carrier all landed in one window: 7,139 hops, 30 of 30 mints and 8 of 8 carriers of one match accepted (combined target met, retry 1) Desk-node accepted hops and mints (coordinator read; Mini read the same per step) and guest carrier ids found among the desk node's accepted ids. Not Pulse's count tn10-storm-2026-11-11-prerun-24-retry-1
13 Nov, pre-run 24 (21:55–22:27 UTC) Covenant hops and a KCC-20 mint line landed together: 6,678 hops and 30 of 30 mints accepted in one window; guest carrier not run, so the combined target was not met (retry 0) Desk-node accepted hops and mints (coordinator read; Mini read the same per step). Not Pulse's count tn10-storm-2026-11-11-prerun-24
14 Nov, pre-run 31 (10:59–11:06 UTC) Same script as pre-run 30: 463 inputs rejected at transient mass 1,000,664 (the node's own number); 462 inputs accepted and landed (SDK mass 280,478). The line is one input wide (met, retry 0) The desk node's own accept or error string for each single submit, then the sender's UTXO set for the two it accepted. Not Pulse's count tn10-storm-2026-11-11-prerun-31
14 Nov, pre-run 32 (11:30–12:01 UTC) Twelve covenant families (the step after eight): 20,377 hops, 30 of 30 mints and 8 of 8 guest carriers in one window; ready mass about one block and landings kept up; nothing fell behind (met, retry 0). Not a test of 3,040 payments a second Desk-node accepted hops and mints (coordinator for the window; chain counter for the minutes) and guest carrier ids among the desk node's accepted ids. Not Pulse's count tn10-storm-2026-11-11-prerun-32
14 Nov, pre-run 33 (12:30–13:13 UTC) Payment ladder next to 12 covenant families, a mint line and one tic-tac-toe match: combined target met (13,620 hops, 30 of 30 mints, 8 of 8 carriers). Payments at 150 sompi/gram included at the offer through 2,027/s; floor hops at 100 left about 950 a minute already at the 1,013 step. At 3,040 and one minute near 4,053, submit-OK stayed near the offer and same-minute inclusion of our payments stayed near 123,000–133,000 a minute. By the end of the tail, 3,272,071 of 3,272,071 payment submits were included (met, retry 0) Desk chain counter by accepting chain block time, and the coordinator for the window. Not Pulse's count tn10-storm-2026-11-11-prerun-33
14 Nov, pre-run 34 (13:44–14:19 UTC) Same ladder with floor families at 200 sompi/gram. HOLD-1 and HOLD-2 met on payment submit-OK. At 1,013, floor hops were 709–778 a minute (quiet 796 and 833). At 2,027 they were 315–381. HOLD-3 did not meet: payment submit-OK fell to about 7,000 a minute with our in-memory pool. RETURN was not scored. The runner stopped at 14:19:36Z on usable tKAS under 18,000. Guest 8 of 8. Mint 90 accepts, 0 rejects (not met, retry 0) Desk chain counter by accepting chain block time, for the published minutes, and the counter's last snapshot at the stop. Not Pulse's count tn10-storm-2026-11-11-prerun-34
14 Nov, pre-run 35 (14:54–15:39 UTC) Same fee as pre-run 34, submitted through RETURN. Combined target met (23,781 hops, 120 mints, 8 of 8). Quiet floor hops 957 and 917. At 2,027, payments stayed near the offer and floor hops were 339–430. At 3,040, submit-OK stayed near 182,400 and same-minute inclusion of our payments was 118,500–126,056. The 4,053 minute submitted 242,003 of 243,180 and included 124,792. RETURN submit-OK returned to about 60,400–61,300; inclusion stayed 126,402–135,121 and floor hops stayed 365–384. HOLD-1 did not meet (our coin line). Usable 17,999 at 15:33:12Z; the runner closed return-closed (met on the three-part target, HOLD-1 not met, retry 0) Desk chain counter by accepting chain block time, for the published minutes, and the target file for the window. Not Pulse's count tn10-storm-2026-11-11-prerun-35
14 Nov, pre-run 36 (15:57–16:09 UTC) Ready cap 8 on our heavy lane, rate 460, no payment ladder and no families. The lane log held 8 pending and about 233–240 submit-OK a minute. The probe's template median was about 266,700–269,700 storage mass, one heavy transaction, and 16 of 17 HOLD samples were under 400,000. The lane was back to pending 0 in the first return minute. 1,421 submit-OK, 0 rejects, fees 1,744.87 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-36
14 Nov, pre-run 37 (16:24–16:36 UTC) Ready cap 16 on our heavy lane, rate 460, no payment ladder and no families. The lane log held 16 pending and 420–449 submit-OK a minute. The probe's template median stayed about 266,700–266,800 storage mass, one heavy transaction, and all 17 HOLD samples were under 400,000. Ready-mass medians were 4.43–8.55 blocks. The fee cap stopped new submits at 16:32:41Z. The lane was back to pending 0 in the first return minute. 2,451 submit-OK, 0 rejects, fees 3,007.72 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-37
14 Nov, pre-run 38 (16:43–16:55 UTC) Ready cap 32 on our heavy lane, thirty chains, rate 460. Pending sat at 30, and submit-OK was 501–540 a minute. While our heavy class was 18–28 and ready, the template's heavy count stayed 1, and 56–576 other transactions sat in the 101–149 fee band. Ready mass in those samples was 17.06–25.72 blocks. The fee cap stopped new submits at 16:50:39Z. Pending 0 from the first return minute. 2,453 submit-OK, 0 rejects, fees 3,010.67 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-38
14 Nov, pre-run 39 (17:04–17:16 UTC) Sixty chains, ready cap held at 32, rate 460. Pending sat at 32, and submit-OK was 523–576 a minute. While our heavy class was 16–32 and ready, the template's heavy count stayed 1, and 133–661 other transactions sat in the 101–149 fee band. Ready mass in those samples was 13.65–28.66 blocks. The fee cap stopped new submits at 17:11:26Z. Pending 0 from the first return minute. 2,447 submit-OK, 0 rejects, fees 3,002.96 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-39
14 Nov, pre-run 40 (17:24–17:36 UTC) Sixty chains, ready cap off, rate 460. Pending reached 60, and submit-OK was 519–631 a minute. While our heavy class was 30–60 and ready, the template's heavy count stayed 1, and 532–1,941 other transactions sat in the 101–149 fee band. Ready mass in those samples was 30.66–49.71 blocks. The fee cap stopped new submits at 17:31:15Z. Pending 0 from the first return minute. 2,445 submit-OK, 0 rejects, fees 3,002.11 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-40
14 Nov, pre-run 41 (17:49–18:01 UTC) One hundred chains, ready cap off, rate 460. Pending reached 100, and submit-OK was 558–658 a minute. While our heavy class was 88–100 and ready, the template's heavy count stayed 1, and 451–3,355 other transactions sat in the 101–149 fee band. Ready mass in those twelve samples was 54.96–76.80 blocks. The fee cap stopped new submits at 17:56:03Z. The 80-block pause did not trip. Pending 0 from the first return minute. 2,448 submit-OK, 0 rejects, fees 3,003.52 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-41
14 Nov, pre-run 42 (18:09–18:21 UTC) One hundred chains, ready cap off, rate 460, fee cap 6,000 tKAS. The six minutes completed. Pending reached 100, and submit-OK was 552–645 a minute. While our heavy class was 84–99 and ready, the template's heavy count stayed 1, and 756–3,025 other transactions sat in the 101–149 fee band. Ready mass in those eighteen samples was 54.24–78.48 blocks. The fee cap did not trip. The 80-block pause did not trip. Pending 0 from the first return minute. 3,576 submit-OK, 0 rejects, fees 4,388.53 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-42
14 Nov, pre-run 43 (18:36–18:48 UTC) One hundred chains, ready cap off, rate 920, fee cap 6,000 tKAS. Pending reached 100, and submit-OK was 449–661 a minute. While our heavy class was 88–99 and ready, the template's heavy count stayed 1, and 623–6,180 other transactions sat in the 101–149 fee band. Ready mass in those twelve samples was 54.01–80.24 blocks. The pause tripped at 18:41:13Z and cleared at 18:41:33Z. The fee cap stopped new submits at 18:43:16Z. Pending 0 from the first return minute. 2,449 submit-OK, 0 rejects, fees 6,012.63 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-43
14 Nov, pre-run 44 (18:57–19:09 UTC) One hundred chains, ready cap off, rate 920, fee cap 12,000 tKAS. The six minutes completed. Pending reached 100, and submit-OK was 450–652 a minute. While our heavy class was 87–99 and ready, the template's heavy count stayed 1, and 506–6,951 other transactions sat in the 101–149 fee band. Ready mass in those sixteen samples was 54.83–84.9 blocks. The pause tripped at 19:02:31Z and at 19:05:13Z, and each clear was the next sample. The fee cap did not trip. Pending 0 from the first return minute. 3,274 submit-OK, 0 rejects, fees 8,041.95 tKAS (met on the table and the zero, retry 0) Desk probe every 20 s, and the lane's own minute log. Not Pulse's count tn10-storm-2026-11-11-prerun-44
14 Nov, pre-run 45 (19:43-20:07 UTC) Tic-tac-toe, covenant hops at 200 and a KCC-20 mint, then the heavy lane at 920. Nine families included and four dropped for no funding, so eight families hopped. Mint 30 of 30 finished in quiet, fees 1.2317 tKAS. Guest 31 matches, 225 submit-OK, 224 accepted on the desk. Matches 1-21 confirmed 8 of 8. Matches 26-30 confirmed 8 of 8 at quotes 1,505.14-1,559.99. While our heavy class was 79-99 and ready, the template heavy count stayed 1 and storage mass was 266,682-310,994. Hops submitted 4,827 OK in quiet and 181 in congest. The fee cap stopped the lane at 8,001.53 tKAS. The lane end line says pending 56. Probe samples after the stop are at fee 100 (met on the table, pending is not zero, retry 0) Desk probe every 20 s, the lane log, the guest log and the family submit log. Not Pulse's count tn10-storm-2026-11-11-prerun-45
14 Nov, pre-run 46 (20:52-21:24 UTC) Tic-tac-toe, covenant hops at 2,000 and a KCC-20 mint beside the heavy lane at 920. Thirteen families included and twelve hopped. Quiet hop submit-OK 5,956, full minutes 582-627. Congest hop submit-OK 3,330, minutes 521-592, including 521-539 while the lane was submitting. Mint 15 of 24, fees 0.6169 tKAS, stopped on the phase file. Guest 33 matches, 264 submit-OK, 264 accepted, each 8 of 8, then a refill refuse-fee stop at 21:14:05Z. Six samples while our heavy class was 86-100 and ready: template heavy count 1, storage mass 266,752-484,443. The pile emptied at 21:07:41Z. Lane 1,376 submit-OK, 0 rejects, fees 3,385.81 tKAS, end line pending 0. Fee cap 8,000 did not trip (met on the table, pending is zero, retry 0) Desk probe every 20 s, the lane log, the guest log, the mint log and the family submit log. Not Pulse's count tn10-storm-2026-11-11-prerun-46
14 Nov, pre-run 47 (22:58-00:11 UTC) Payment ladder at 100 sompi/gram with a 50 slice and a 1,000 slice, twelve covenant families, one KCC-20 mint, and tic-tac-toe. No heavy lane. HOLD-1 full minutes 60,307-61,367 against 60,780. HOLD-2 120,805-122,485 against 121,620. HOLD-3 through 23:32Z 180,640-183,137 against 182,400, then 150,046, 17,383, 11,094 and 9,906. After a pool reload at 23:41:26Z, 23:42 and 23:43 are 242,501 and 242,413 against 243,180. Six senders stop reason=rate. The 50 slice has 0 submit-OK. The 1,000 slice sits on 590-600 a minute through 23:30Z. Families 1-13 included. Family submit rows stop after 23:04Z. Twelve STALL lines at 23:05:25Z, and 36 STALL lines in the run. Mint 20 of 20, fees 0.82 tKAS. Guest in the locked window: 114 matches, 902 submit-OK, 902 accepted. 111 matches are 8 of 8. Matches 38, 42 and 74 are 3, 4 and 7. Guest stop 00:11:18Z. RETURN was not written. Pay-100 minute fee fields 8,675.68 tKAS over 4,668,093 submit-OK. The 1,000 slice 322.13 tKAS over 19,061. The 50 slice 0 (met on the table, retry 0) Payment minute logs, the guest log, the mint log, the family submit log and the desk probe. Not Pulse's count tn10-storm-2026-run-47
15 Nov, pre-run 48 (11:22-15:50 UTC) Three vprog kinds, then one payment process at 1,013, 2,027, 3,040 and 4,053, fee 100, no families. Locked-ladder full minutes: 1,013 step 60,179-61,336. The 2,027, 3,040 and 4,053 steps are 71,154 or 72,360. Those three offers were not reached. Zero rejects. A later one-process sender that reuses a coin in the same second: 103,716-106,128 on 2,027 and 102,510-104,922 on 4,053. The 3,040 step is not one band (83,214-106,128). Quiet game 24 of 24. Fresh-directory batch 192 of 192 during the 1,013 step. Reused directories 0 accepted on the desk. Desk notes: account and flow completed a pass at quote and at 100. Flow at 2,000 logged no fundable transaction. Account at 2,000 did not confirm one lane action (met on the table, retry 0) The payment minute log and the desk notes. Not Pulse's count tn10-storm-2026-run-48
15 Nov, pre-run 49 (19:17-20:23 UTC) Two payment processes, disjoint halves of 1,206 coins, fee 100, each offering 1,520 a second. Five full minutes of combined submit-OK: 183,530, 190,827, 189,662, 181,172 and 181,032. All five are at least 2,200 a second. Three are over 182,400 and two are under. Zero rejects. Mempool field 173,092 to 905,883. Stopped at 19:22:33Z on memory. The 4,053 offer was not raised. The engine was down during the payment window. After it, one tic-tac-toe match accepted 8 of 8 at quote 216.16. Account at 100 completed a scripted pass while sync was behind. Account at 2,000 completed a scripted pass at a 20-second step after two shorter starts did not. Flow at 2,000 submitted 2 activity transactions. Receipts are dev-mode and are not real ZK receipts. Submit-OK is not inclusion. Last run of this stretch. On 15 Nov 2026 we call it a day until further notice (met on the table, retry 0) The payment minute log and the kind phase lines. Not Pulse's count tn10-storm-2026-run-49

11 Oct in one line: network unique accepted, 07:30-12:33Z (303 minutes, halted early by stp: taking a break, other things on the agenda), about 51.5 million unique transactions; peak complete minute 3,583.1 tx/s at 11:28Z, mean 2,833.1, median 3,107.1 over 171 strict single minutes; includes about 56 tx/s base load. Independent counts by Kaspa Pulse (@gokugalax, kaspapulse.com): accepted in block counts on two external nodes, best minute 3,646, 51.2M accepted (07:30-12:30Z). Build's results: pending. Source: logs/bot/SUMMARY.md, logs/COMBINED.md.

Confirmation, Sun 8 Nov 2026, 12:00 UTC, finished 16:10 UTC: tn10-storm-2026-11-08-confirm. Same shape as the 00:40 rehearsal, on its own clock. Both publishers fired. Desk load mean 3,006.8 over 2,700 s, under the 3,078 line. Sixteen mean 1,111.1 over 900 s, every family above 1,026. Mini's three files are a separate reading. The method log is logs/METHOD.md. Not a restart of the 00:40 rehearsal. Not the counted storm.

Named count, Tue 10 Nov 2026, 19:00 UTC: tn10-storm-2026-11-10. Pulse named this day. Same shape and times as Saturday 7 Nov. The operator sends the pastes on Tuesday. A read before that day does not start a runner. File 1 is due Wed 11 Nov at 12:00 CET. File 2 is due Thu 12 Nov at 18:00 UTC. Wednesday is not a spare rerun of this clock. Desk paste GO-DESK.md. Mini paste GO-MINI.md.

Wed 11 Nov 2026, 12:00 UTC, not armed: tn10-storm-2026-11-11. Sixteen covenant families under a payment climb: 1,013, then 2,027, then 3,040, then one observation minute at 4,053, then 1,013 again. A scored step closes on three clean minutes of submit-OK, or as NOT MET after one repair pass. There is no latest minute after the start. The desk sends the families and every payment offer. Mini counts. The bot sends payments only in the 4,053 minute. The pastes refuse to start until REVIEW_CLOSED. Desk paste GO-DESK.md. Mini paste GO-MINI.md. Bot paste GO-BOT.md. The rehearsal of this plan is in the pre-run repos; it did not arm this neuron and did not write REVIEW_CLOSED.

11–13 Nov rehearsal (pre-runs 1–23): pre-run 1 (tn10-storm-2026-11-11-prerun) opened at 23:05 UTC on 10 Nov, then failed before hashes after the desk and Mini reset. The series continued in tn10-storm-2026-11-11-prerun-2 through prerun-23. Pre-runs 7–11 met the scored submit-OK steps; pre-run 12 was observation only; pre-run 13 was observation only and failed before RETURN on a coin-count abort; pre-runs 14, 15 and 16 were observation only and finished through RETURN; pre-run 17 was observation only and failed before RETURN on its own abort line (key 0 refilled too little); pre-runs 18 and 19 were observation only and finished through RETURN; inclusion is a separate desk-node reading, and Kaspa Pulse has not counted it. In pre-run 10, cheap chained tiers were already slow at the 1,013 baseline, unlike pre-run 9; in pre-run 11 we observed that the floor tier followed the ready mass in our node's mempool, driven mostly by another sender's heavy stream at the same fee, more than our offer; in pre-run 12, with no load of ours, fee rates near the floor (100 to 200) only separated while that stream was present, and by much less than the selection weights, and our SDK's storage-mass helper read about four times low for that stream's shape; in pre-run 13, with our own heavy lane as the only competitor (operator's OK, stated publicly because it slowed other users' low-fee transactions), every tier slowed in steps as the lane's ready count rose and recovered within a minute, and the node's own fee rate for our hops was about a third of the built rate, which corrects the "same fee" reading of earlier pre-runs; in pre-run 14, with the hops priced on the node's own mass, a hop at our lane's rate kept 76–88 percent of its BASE rate, about twice the lane's rate kept it within about 4 percent, and the families recovered within a minute (our lane again slowed other users' traffic, stated publicly); in pre-run 15, with our lane uncapped at 126 and 150, about twice the lane's rate kept 95–98 percent and about three times was needed at the heaviest step; in pre-run 16, a fee-escalation duel with both sides ours settled near twice the lane's rate, with no spiral (our lane again slowed other users' traffic, stated publicly); in pre-run 17, two equal heavy lanes of ours on one rule stepped each other down to 100 in four minutes, no spiral; in pre-run 18, with one of the two lanes greedy, we observed a leapfrog climb from 150 to 1,405 in about 21 minutes, no cap, the lanes landing no more together and the fixed tiers squeezed near zero (our lanes slowed other users' traffic by about 23 percent on our node at the top, stated publicly); in pre-run 19, the control, a greedy lane alone climbed too (150 → 719 until its spend cap) with flat landings, a ceiling rather than competition, and still squeezed the fixed tiers (our lane slowed other users' traffic by about 5 percent, stated publicly); in pre-run 20 a never-down greedy rule ratcheted even with a reachable target (150 → 368) while the normal rule settled at 100, and at a fixed 460 the desk node's template stayed full with 8 chains of ours and shrank with 30 (our lane slowed other users' traffic by up to about 7 percent, stated publicly). The bot wrote a legal NO SEND (no-free-node) from pre-run 5 on. Added 13 Nov: pre-run 21 was void (a network outage, no usable data, no findings); in pre-run 22 our cap path ran live as planned and paying more did not land more with 8 chains; in pre-run 23 we observed the desk node's template follow how many of ours were ready, not the chain count or the rate (likely, not proven); in pre-run 24 (prerun-24) four covenant families and a KCC-20 mint line landed together (30 of 30 mints), the guest carrier was not run, and the combined target was not met. The log is FINDINGS.md items 22–82, the history page is history/storms/2026-11-11-prerun.md, and the current playbook is PLAYBOOK.md. This rehearsal did not arm Wednesday and did not write REVIEW_CLOSED.

13 Nov 2026, public TN10 node count (a check, not a run): a dashboard tile showed 43 public nodes; we found 2 distinct nodes behind the public names, 9 public wRPC nodes (8 synced) and about 39 plausible P2P-reachable nodes. Our guess, not confirmed, is that the 43 counts P2P-reachable nodes. Limits and method: FINDINGS.md item 80. Corrections welcome.

Mint rush, plan only (no date yet): tn10-storm-plan-mint-rush: a KCC-20 mint rush under storm load, capped auto-fee vs fixed-fee minters. Hypothesis: stp's; test idea: TN10 ops (the bot). It stays its own map.

8 Nov, the rehearsal: tn10-storm-2026-11-08. It stands on its own. The desk night closed at 04:50 UTC. The scores are one desk reading. Mini's three count files are in that repo. LESSONS.

Map

tn-master/
  README.md                 this page
  WHAT-WHY-HOW.md           what we test, why, how
  METRICS.md                TPS vs tx/s: one vocabulary, Kaspa Pulse's counter, base load
  FINDINGS.md               numbered findings, each with a label and a source
  PLAYBOOK.md               current pre-run playbook (from pre-run 9, plus lessons)
  ADVICE-TO-DEVELOPERS.md   what we would tell a builder, from what we saw
  history/                  MASTER.md (every storm), TIMELINE.md, storms/ (one page per storm)
  credits/                  KASPA-PULSE.md (every input and what we changed), kaspa-pulse/ (his counts)
  repos/                    INDEX.md (every storm repo), RENAME-LOG.md (old -> new names)
  docs/                     NEURON.md, LESSONS.md, OPEN-QUESTIONS.md
Page What
WHAT-WHY-HOW.md What a storm is, why we run them, how (nodes, senders, clock, counting)
METRICS.md One vocabulary for tx/s; Kaspa Pulse's "accepted in blocks" counter; base load
FINDINGS.md Numbered findings with labels and sources
PLAYBOOK.md Current playbook for the 11.11 rehearsal and later pre-runs
ADVICE-TO-DEVELOPERS.md Fees, re-broadcast, coin selection, measuring
history/MASTER.md Every storm, oldest first (moved here from tn10-storm-2026-10-10)
history/TIMELINE.md One line per day
history/storms/ One page per storm
credits/KASPA-PULSE.md Kaspa Pulse: every input and what we changed
repos/INDEX.md Every storm repo: name, old name, date, class, visibility (incl. the mint-rush plan)
repos/RENAME-LOG.md The 11 Oct 2026 renames
docs/NEURON.md Every storm repo is indexed here the day it is created, and locked here the hour the clock closes
docs/LESSONS.md · docs/OPEN-QUESTIONS.md Operations lessons; what still needs testing
tn10-lowio-1148-notes Notes on rusty-kaspa PR #1148 (lowio preset): diff read, risks, A/B test (FINDINGS 18)

Counts only. No IP, host, port, key, seed, txid or wallet address on these pages. Public TN10 node names are allowed.


About

Map of every Kaspa TN10 storm (testnet only). Thanks to Kaspa Pulse (@gokugalax).

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors