igneum-labs
|
c43880ef41
|
Class v6 10.0j (after the landing): the multi-family adversary lane's macro-window core (k 0.40 node-for-node; the window knob buys the card nothing against a macro file) carried beside the k lane's model as a disagreement the placed rows settle
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 15:13:27 +00:00 |
|
igneum-labs
|
7e5897dc32
|
Class v6 10.0i: the adversary's re-optimised core (the gated base k 0.37 at N3, the gated window 0.50, +0.13 of k the window's honest defence; the placed ungated core +64 percent over synthesis) and the bracket the served figures sit in until the placed gated rows at 17:30
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 15:04:41 +00:00 |
|
igneum-labs
|
e1d397e8b0
|
Merge the mirror's master da5e1e8a into counter-asic-4 (the coord record lane's earlier copy of class-v6-rotating-family.md superseded by this branch's)
# Conflicts:
# docs/design/class-v6-rotating-family.md
|
2026-10-08 15:03:33 +00:00 |
|
igneum-labs
|
01d2848791
|
Class v6: the window's card cost MEASURED (within 5 percent per load with the liveness chain, no spill); the served sentence's one word made honest (adopts, retains across every rotation); lane D's report (x16 over the verifier gate; the genesis table's 75-point spelling under test); lane C's drawn-era F8 read (the per-load class must take (c''')); lane 1's memory-clock ladder (the memory side bought one for one with the rate)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:56:22 +00:00 |
|
igneum-labs
|
698837d70a
|
Class v6: the liveness rule on the window in change (4) (main, 15:56 UK); the three v7 research lanes named
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:42:27 +00:00 |
|
igneum-labs
|
b6a45ced76
|
Class v6 10.0f: lane 3's profitability surface, first cut (p* scales as C_dev over q times the discounted life and under 5 percent with the per-joule edge; the conditions under which development is attractive; the fixed-lane chip needs 4x the price of the programmable one); the served economic line carries it
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:35:20 +00:00 |
|
igneum-labs
|
6211413a74
|
Class v6 10.0h: the served text (the review's sentence verbatim, each number's label, the lines beside it, what is withdrawn)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:32:53 +00:00 |
|
igneum-labs
|
15d1c284fa
|
Class v6 10.0g: the second external review's seven amendments (energy, economic and response statements kept separate; the minimax evaluation rule with the free adversary and the negative controls; the discrete-GPU cohort; the 1.25x same-node gate; no arrival probabilities; the RandomX wording; the next programme named)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:31:49 +00:00 |
|
igneum-labs
|
ad00108853
|
Class v6 10.0f: the external review's five corrections (the programmable chip's 3-year life with the per-life table, the lifetime headline withdrawn; the profitability surface in place of the single threshold; the dataset claim narrowed to a stale machine; proving an opportunity not an exclusion; the served sentence as the review words it; Monero an observed comparison; the window k not a lower bound); the 32-lane window core as the chip assessed (about 2.6x a node ahead, 2.2x node-for-node)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:30:45 +00:00 |
|
igneum-labs
|
44eacd733d
|
Class v6 amendment 15:3x UK: the chip's shuffle is k 0.011 to 0.021 (the lowest family), the register file linear in its entries (the window is the knob), the core32 row; change (1) amended: the op mix moves to the band's best mix as the genesis table, subject to one acceptance pass
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:28:58 +00:00 |
|
igneum-labs
|
731caa1f18
|
Class v6 3.3: lane 3's reading of the measured 5.5 GiB step (the mapping safe at the v6 epoch; 4 percent at stock, about 9 at the knee; the 8 GB tier's RX 7600 row owed)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:25:50 +00:00 |
|
igneum-labs
|
a41aa5a0cc
|
Class v6 3.3: the 5.5 GiB genesis floor measured on a rented 5090 at stock (3.5 percent of rate, about 4 percent of energy; the non-power-of-two mapping no cliff on sm_120; the ds55 pin)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:24:53 +00:00 |
|
igneum-labs
|
612764d819
|
Miner window audit: the Earnings row is two windows (7 days, all time) under the 24-hour headline; the 24-hour count never falls back to zero before the engine's earned sums land
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:23:33 +00:00 |
|
igneum-labs
|
07cb5d933a
|
Class v6: W = 8 withdrawn on measurement (the 5090 at stock pays +4.8 percent of energy for 0.2x of the die's shadowed edge; lane 3 083ea1e4); the width is W = 4, 16 never
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:15:53 +00:00 |
|
igneum-labs
|
9ff09ecc9f
|
Class v6: W = 16 closed from the knee side (2.3x the energy per hash at the 5090's lock); nothing on it owed
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:05:47 +00:00 |
|
igneum-labs
|
86a9f4f27a
|
Class v6 layer 1: kit d measured (inside the shadow block the weight table is a 30 percent energy lever; multiply-heavy 23 to 33 percent under shuffle-heavy per hash; shuffle-heavy = stock in watts; the lossy cap stands on both sides); W = 16 dead at the knee too
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 14:04:44 +00:00 |
|
igneum-labs
|
1a8629c758
|
Class v6 10.0c: the 32-lane 16-register row (k 0.34 at the lock at N3: wider SIMD amortises the sequencer; the window stays the only knob that does not)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:55:45 +00:00 |
|
igneum-labs
|
8d6c638616
|
Class v6 10.0e: the node sentence (2.0x own node, 2.4x a node ahead, 2.8x two ahead; the honest tier moves nodes with every GPU generation, a chip re-tapes-out)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:54:21 +00:00 |
|
igneum-labs
|
d9a08bc753
|
Class v6 10.0e: the node column (node-for-node the window core is k 1.09 and the DRAM board 2.0x at the 5090's knee; N3 buys the chip 0.4x, N2 0.8x)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:53:36 +00:00 |
|
igneum-labs
|
1f1a3b361e
|
Class v6 10.0e: the close with the 64-register window (the GPU-tier floor 2.2x to 2.4x per joule at the knee, under 2x only against the Apple tier at 1.5x; the card's window cost modelled, labelled; Monero's 1.0x to 1.5x beside it); the served sentence re-folded
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:52:07 +00:00 |
|
igneum-labs
|
e5a72c5d82
|
Class v6 10.0c: the k lane's design sweep (k 0.85 not reached by any amortisable knob; the 64-register window the one robust knob at k 0.78; the long block 0.97 only on a flop-array imem); the close's fourth change
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:50:37 +00:00 |
|
igneum-labs
|
beac74a823
|
Miner window audit: the dev fee section is last on Settings with no control (the coordinator's order of 8 October 2026, earnings-26); T-R12 counts s-fee-card and asserts no input
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:50:17 +00:00 |
|
igneum-labs
|
35f89908d0
|
Miner window audit, 8 October 2026: every page as it stands, the register of 21 repeats and 21 misleading readouts with a verdict each, the four-page layout (Mine, Tune, Earnings, Settings) row by row with the words, the release-0.3.26 build plan for the UI lane (8 items, hours, one test per repeat), five questions held for the founder with defaults; the record joins the export exclude list
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:50:17 +00:00 |
|
igneum-labs
|
41670ab962
|
Class v6 10.1: floor lane 1's close row (the SM count is not the lever on three cards; the residual as the awake floor plus the GPU-side memory path per read; the fraction a chip cannot strip; what ships; the memory-clock ladder owed)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:49:12 +00:00 |
|
igneum-labs
|
b53c561e5a
|
Class v6 10.0d: the rotation schedule the close adopts (hourly 8,766 / weekly 52.18 / family 2.03 / vote at most 2.03 extra; the detector's false positives on today's rows; the 6 h vote window; the rotation lane bd43f808)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:46:25 +00:00 |
|
igneum-labs
|
a28ea58f9b
|
Class v6 layer 4: lane D's measured resolution of the bucket bound (retired into one value-level index-bit test; seven known-failed cases of the fold; 300 sigma if a refusal is wanted before the fold) and of the width-4 floor (stays 0.995 at its measured cost)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:45:00 +00:00 |
|
igneum-labs
|
b114f24599
|
Class v6: floor lane 4's final table (the class v5 stock column measured on 14 card classes; the lock worth 34 to 41 percent on Blackwell and Ada; the 16 GB Blackwell knee the honest floor); layer 1's N row carries the measured price of a 4x shadow (1.58x the energy per hash on the 5090 and 4090)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:42:50 +00:00 |
|
igneum-labs
|
22a3ba3c50
|
Class v6 10.4: rented 5090 and 4090 stock rows on class v5 genesis (3.14 and 4.48 microjoules) and the measured price of a 1.5x heavier shadow (1.58x the energy per hash on both architectures; the 1p5x-knobs lane)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:41:12 +00:00 |
|
igneum-labs
|
784a8dbd35
|
Class v6 layer 5: lane C's document landed on master at bfe7607a; the F8-form no-era read clean on 59 of 60 seeds
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:12:35 +00:00 |
|
igneum-labs
|
161bb535e4
|
Class v6 10.0b: the precedent row replaced by floor lane 5's sourced pass (RandomX chipped at 46 months at 1.0x to 1.5x by the Antminer X5, correcting the chip-free premise; Ethash 36 months, 14x today; Kaspa 21 months, 167x to 725x)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:10:22 +00:00 |
|
igneum-labs
|
9c2ae21068
|
Class v6: the Kaspa precedent with the sourced 4090 row (170x the first chip, 730x the 2024 one); the census lane's acceptance rows for W = 16 and the re-weight on layer 1 (clean on acceptance; decided on energy)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:08:46 +00:00 |
|
igneum-labs
|
d1170a2ef5
|
Class v6 10.0a to 10.0c: cost per hash over the chip's 180-day life (under 1x only above about USD 300 M a year of miner revenue; the three-price table), the precedents sourced (RandomX design document sections 1 and 3.4, no chip in 70 months; Ethash 13x at 36 months; Kaspa hundreds of times at 20 months), the core-shape rule for the k lane's sweep; the served sentence in two units
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:07:45 +00:00 |
|
igneum-labs
|
c5f7125348
|
Class v6 10.3: the SRAM rows on the synthesised core (lane 3 b2563b78: 8.1x stock, 5.6x knee at class v4's shadow; 3.5x and 2.2x at the full shadow); 10.2 the net of the two corrections per node
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:05:05 +00:00 |
|
igneum-labs
|
ce1ccfb8ce
|
Class v6: lane C's drawn-era re-read of the sound per-load form (254 of 256 at 0.819; the 16 x 27 form dead on both instruments; eras spread the bit-0 refusals); the hash lane's two F8-256 attributions on layer 4's bucket-bound row (both tests carried, labelled estimate)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:04:00 +00:00 |
|
igneum-labs
|
efdcf726c1
|
Class v6: the k lane's sequencer-core row is the headline (k 0.56 at the lock, 0.31 at stock, 0.50 against the M5 Max, absolute, N3, synthesis-only; the DRAM chip 2.8x at the knee and 4.1x at stock, inside the record's claimed band); the close table, the lead, the one-line floor and the served sentence re-folded on it
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 13:02:58 +00:00 |
|
igneum-labs
|
c1bf30676d
|
Class v6 10.3: the W = 16 net on the 5090's measured 1.25x (lane 3 bbf1b1dc); the verdict stands on four cards
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 12:40:23 +00:00 |
|
igneum-labs
|
56276ff039
|
Class v6 10.5: the 5090 W = 16 row measured (the hinted sparse form holds the rate for +88 W, energy +25 percent; the verdict holds on every card)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 12:38:48 +00:00 |
|
igneum-labs
|
9e603e3af6
|
Class v6 10.3: W = 16 netted on measured energy (lane 3 a670ba28): the SRAM die lands where W = 8 puts it for free with the shadow on; W = 8 is the width, never 16
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 12:31:53 +00:00 |
|
igneum-labs
|
51ac13a43b
|
Class v6 10.5: W = 16 killed at stock on measured rows (the 4090, H100 and 3090: the hinted form holds the rate on Ada, the card pays the second sector at 8 to 9 nJ, energy per hash +34 percent against the chip's +33; k about 0.13)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 12:30:37 +00:00 |
|
igneum-labs
|
aab62cad14
|
Class v6: the last two 20:00 and 19:30 references pulled to the close
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
|
2026-10-08 12:28:29 +00:00 |
|
igneum-labs
|
270aebfc9b
|
Class v6: the last two 20:00 and 19:30 references pulled to the close
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 086411627 (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
aae3ef5246
|
Class v6: the coordinator's 13:30 UK clocks (floor rows by 15:30, the synthesis on master by 16:30; the decision stays the founder's)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 931798ead (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
ebe0467309
|
Class v6 10.1: the 5090 SM ladder on four rented hosts (the ceiling held to 43 of 170 SMs, the occupancy 1.3 to 3.9 percent at best), the H100 power decomposition, the fraction a chip cannot strip per tier
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of f21f256a0 (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
105fa65d5e
|
Class v6 10.4: the honest denominator per tier (floor lane 4: the 16 GB Blackwell card at its knee is the honest NVIDIA floor, the 5080 2.06 measured; the only lever the operating point; the Ember tier table); the close table gains the 5080 row
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 9629fd8db (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
f2b13d062f
|
Class v6: 15:45 UK is the close (the founder's clock rule); later rows are amendments with their own time; no 20:00 event
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of b351f3fdf (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
1cbf16eb37
|
Class v6: 10.1 SM-sparse measured dead on the 4090, H100 and 5090 (the residual is the clock domain); lane 3's close rows (the capex wall on the corrected project floor: USD 23 M and 340 M a year; W = 16 per chip; the SRAM rows on the synthesised core); 10.0 carries them with the 4090 and H100 rows
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 00890f9cc (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
cbaa34c9ff
|
Class v6 10.0: the provisional close frame for 15:45 UK (one table, honest tier per row, chip per column, default k 0.5 and the floor k per cell; the capex wall's two thresholds; the served line in one sentence; the three class v6 changes), every missing row's default stated
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of f0146c2c2 (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:21 +00:00 |
|
igneum-labs
|
d4d2ac5791
|
Class v6: 10.5 invention lane's KEEP/KILL table (the second sector conditional on the w64 job; everything else killed with its number); 10.2 reframed as per-unit floors with the sequencer-core row owed as the headline and k 0.5 the default; lane D's complete lossy curve and ring-C progress in 5.1b
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 89ea64861 (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:20 +00:00 |
|
igneum-labs
|
62ccca7d3b
|
Class v6 10.2: the shadow's k from RTL (floor lane 2: k 0.18 at the lock at N3 absolute, a floor; the GDDR7 chip with the class v4 shadow 4.0x at the knee and 5.8x at stock against the served 2.1x; mulhi the worst lever on the card by 6x); the lead and the served-line review carry it
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of 4e7bac031 (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:20 +00:00 |
|
igneum-labs
|
13a8ae3159
|
Class v6 layer 4: lane D's 17:00 cut as 5.1b (24,000 drawn eras: the op-mix band's known-failed test reads 0 of 3,000 under the cap; the lossy curve's edge at +2 to +3; the (c''') per-width calibration; the bit-R bias law; the measured bound n)
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Documents-only replay of dd692978a (counter-asic-4) for the box mirror master
|
2026-10-08 12:28:20 +00:00 |
|