/download and /miner: the fresh-sync memory blocker stated as the real minimum (64 GB today, 48 GB the floor at this hour, growing with the chain until the snapshot path 2.0.3.1 ships; the measured growth and the 16 GB and 32 GB outcomes in plain words); the class v4 pairing lines that go whatever the X0 amendment says (the 5 October recompute-chip line on the litepaper and /claims, the latency-shadow clauses, the ledger's 3.0x to 3.2x figure); the devnet runs class v5 from its first block per the manifest

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

docs/fud-ledger-2.0.md

@@ -22,7 +22,7 @@ Pin: Architecture and product, sixth box (every served page carries the position
 
 Status: Open (8 October 2026): pass when the best supported cost and energy advantage of a re-optimised, programmable adversary sits inside the chosen competitiveness envelope, with its uncertainty published (D3).
 
-Answer: The target is contestable mining, not chip destruction: a manufacturer with a profitable product is not the failure; an exclusive, durable advantage large enough to displace the accessible GPU fleet is. Two measured results already bound the design. The connected-state reorganisation was killed as a class because rearranging the same operations moves neither side (D2). The mixed FP32 branch was killed because deterministic FP32 costs the cards 15 to 26 percent energy per hash against a 10 percent budget and widens the chip's edge to 3.0x to 3.2x (measured cards, modelled chip). Both stay as regression controls.
+Answer: The target is contestable mining, not chip destruction: a manufacturer with a profitable product is not the failure; an exclusive, durable advantage large enough to displace the accessible GPU fleet is. Two measured results already bound the design. The connected-state reorganisation was killed as a class because rearranging the same operations moves neither side (D2). The mixed FP32 branch was killed because deterministic FP32 costs the cards 15 to 26 percent energy per hash against a 10 percent budget and widens the chip's edge (measured cards, modelled chip; the figure paired against class v4 is withdrawn). Both stay as regression controls.
 
 Evidence: `docs/plans/igneum-2.0.md` (the objective, property 1; D1, the mixed FP32 kill; D2(a)); `docs/analysis/class-v6/connected-state.md`.
 

site/claims.html

@@ -249,7 +249,7 @@
 <div class="derived"><p>Here are the limits, stated before anyone else states them.</p>
         <ul>
           <li><strong>A proof in seconds.</strong> Not at launch. Proving a full block today needs a cluster of 100 to 200 consumer GPUs, approximate, so Igneum launches with proofs within about a minute and tightens as hardware improves. Users still see their transaction land in one second.</li>
-          <li><strong>A chip is impossible.</strong> No. Igneum assumes a chip exists. A programmable chip is not stopped by the moving target: everything it needs is public at genesis and every drawn parameter is firmware to it (an address permute, a rotator, an immediate table), so the defence against it is the latency-shadow work (class v4) and the price per joule, not the schedule (the chip and economy analysis of 6 October 2026, section 5.4; <a href="/ledger#D3">ledger D3</a>). Igneum remains competitive on accessible commodity GPUs even when specialised mining hardware is assumed to exist, remain compatible and seek profit; its security does not rely on identifying that hardware or retiring it through emergency changes. Class v6 adopts the 64-register window and retains it across every rotation. Current modelling estimates a 1.5x to 3.1x energy-efficiency advantage for the specialised designs assessed as complete machines against the GPU tier, from a board on commodity DRAM at 1.5x to an SRAM-store die at 3.1x (1.5x to 2.3x on the GPU's own node). For the board on commodity DRAM the two independent chip models agree after placement: 1.5x to 2.1x as a complete machine and 2.0x to 2.9x for the board alone on the GPU’s own node; a node ahead 1.8x to 2.4x and 2.45x to 3.3x (placed and routed on ASAP7, both lanes, the node scaling claimed, the machine terms modelled; the 2.9x end is the resizer-downsized floor of the placed 32-lane comparator, not a point, until the placed core32 lands). The standard’s P04 target reads FAIL at both ends, served as such. The standard’s P04 target as approved: R_E at most 1.5 on the same node and one node ahead; today’s placed bracket reads against it as a FAIL to work against. The long-program and select-tree proposals were rejected. Economic resistance depends on development cost, deployment economics and productive hardware lifetime; family transitions receive an obsolescence benefit only where a loss of competitiveness is demonstrated; programmable multi-epoch designs are included in the assessment. The labels: the bracket modelled, approximate and provisional (the GPU side measured on the RTX 5080 and RTX 5090 at their core locks under class v4, 8 October 2026; the chip core synthesised on ASAP7 and scaled to N3, claimed, its placed gated row pending; its memory modelled). Class v5 makes the dataset the chain’s own state, so a chip that stores it or recomputes it is wrong on every item (designed, 7 October 2026; under evaluation (Deliverable 3), not counted as a defence until justified or dropped). The strongest recompute chip we can price, holding the whole 256 MiB cache on-die, reaches under 1x per chip against an RTX 5090 (the published model, 5 October 2026: 0.92x per unit of silicon with a 3x fixed-function allowance, approximate). Sources: the class v6 close, section 10 (8 October 2026); the ASIC history’s Ethash rows; the chip model analysis (6 October 2026). No hash has stayed free of chips forever; Igneum does not claim to. Monero’s RandomX has held its miners on commodity hardware for about seven years: one chip shipped against it, Bitmain’s Antminer X5 (September 2023), an observed comparison, not a ceiling; the one announced beyond it, the Antminer X9, was withdrawn in mid-May 2026 before any unit shipped, its claimed core never measured; RandomX v2 was released on 25 March 2026 with its activation pending. That record says nothing about the price of a chip with the 256 MB cache on its die; that price is a cost model, not a measurement.</li>
+          <li><strong>A chip is impossible.</strong> No. Igneum assumes a chip exists. A programmable chip is not stopped by the moving target: everything it needs is public at genesis and every drawn parameter is firmware to it (an address permute, a rotator, an immediate table), so the defence against it is the price per joule, not the schedule (the chip and economy analysis of 6 October 2026, section 5.4; <a href="/ledger#D3">ledger D3</a>). Igneum remains competitive on accessible commodity GPUs even when specialised mining hardware is assumed to exist, remain compatible and seek profit; its security does not rely on identifying that hardware or retiring it through emergency changes. Class v6 adopts the 64-register window and retains it across every rotation. Current modelling estimates a 1.5x to 3.1x energy-efficiency advantage for the specialised designs assessed as complete machines against the GPU tier, from a board on commodity DRAM at 1.5x to an SRAM-store die at 3.1x (1.5x to 2.3x on the GPU's own node). For the board on commodity DRAM the two independent chip models agree after placement: 1.5x to 2.1x as a complete machine and 2.0x to 2.9x for the board alone on the GPU’s own node; a node ahead 1.8x to 2.4x and 2.45x to 3.3x (placed and routed on ASAP7, both lanes, the node scaling claimed, the machine terms modelled; the 2.9x end is the resizer-downsized floor of the placed 32-lane comparator, not a point, until the placed core32 lands). The standard’s P04 target reads FAIL at both ends, served as such. The standard’s P04 target as approved: R_E at most 1.5 on the same node and one node ahead; today’s placed bracket reads against it as a FAIL to work against. The long-program and select-tree proposals were rejected. Economic resistance depends on development cost, deployment economics and productive hardware lifetime; family transitions receive an obsolescence benefit only where a loss of competitiveness is demonstrated; programmable multi-epoch designs are included in the assessment. The labels: the bracket modelled, approximate and provisional (the GPU side measured on the RTX 5080 and RTX 5090 at their core locks under class v4, 8 October 2026; the chip core synthesised on ASAP7 and scaled to N3, claimed, its placed gated row pending; its memory modelled). Class v5 makes the dataset the chain’s own state, so a chip that stores it or recomputes it is wrong on every item (designed, 7 October 2026; under evaluation (Deliverable 3), not counted as a defence until justified or dropped). No hash has stayed free of chips forever; Igneum does not claim to. Monero’s RandomX has held its miners on commodity hardware for about seven years: one chip shipped against it, Bitmain’s Antminer X5 (September 2023), an observed comparison, not a ceiling; the one announced beyond it, the Antminer X9, was withdrawn in mid-May 2026 before any unit shipped, its claimed core never measured; RandomX v2 was released on 25 March 2026 with its activation pending. That record says nothing about the price of a chip with the 256 MB cache on its die; that price is a cost model, not a measurement.</li>
           <li><strong>A guaranteed income floor.</strong> No. External proving is a small market today. Igneum's miners' electricity cost in it is close to power, but the price they must charge is the subsidy they forgo, which falls as one over network hash: an edge at scale and nothing more.</li>
           <li><strong>A memory-hard prototype on every vendor.</strong> Not yet. The 256 MB cache closed the shortcut on Apple silicon (computing items runs 4.8x slower than loading them, measured 3 October 2026). The same ratio on NVIDIA and on a discrete AMD card is Open.</li>
           <li><strong>Finality in the first month.</strong> No. No checkpoint locks until the 30-day window has 30 days of history. The first month of mainnet is proof of work with a 12-hour depth, and the text above says so wherever a day count appears.</li>

site/download.html

@@ -282,7 +282,7 @@
       <div class="home-row"><span class="n">03</span><div><b>Press Start, then tell us.</b><p>Watch your blocks ring on the chain. When something breaks, say so on <a href="https://discord.gg/igneum" rel="noopener">the Discord</a> or by email.</p></div></div>
     </div>
     <p class="yourcard" data-yourcard style="margin-top:16px">Your card: pending</p>
-    <p class="home-quiet" data-devnet-reset-note><b>The devnet is being reset this morning, 9 October.</b> A fresh install syncs once the reset completes. This line is re-read at 10:30 UK.</p>
+    <p class="home-quiet" data-devnet-reset-note><b>Before you install: a fresh sync from genesis needs 64 GB of memory today.</b> 48 GB is the floor at this hour, and the number grows with the chain until the snapshot path ships (a fresh node reads about 30 GB of memory by DAA 44,000 and 35 to 40 GB at today’s tip, a gigabyte more every 17 minutes at one block a second; a 16 GB machine reaches DAA 7,000 to 8,000 about four minutes in and then swaps or is killed, a 32 GB machine about DAA 20,000). A smaller machine waits for the snapshot path, 2.0.3.1: its design is due by 12:30 today and the first served snapshot by 15:00; this note changes when it ships. The devnet was reset this morning, 9 October; a fresh install syncs from the reset chain.</p>
     <p class="home-quiet" data-release-note><b>What changed in 2.0.2 (Mac, 9 October).</b> The public miner ships without remote execution; the fleet runs Igneum Miner Lab under its own signing root. Automatic updates are a choice at install and honoured. The recovery lock is named as recovery. A fresh install syncs through the epoch cuts and the node serves its proof records to every joiner. Windows and HiveOS stay on 2.0.1 until their builds are cut; 2.0.1 is the lowest supported version. DMG <code>Igneum-Miner-2.0.2-5d53a591-1176efb9.dmg</code>, sha256 <code>66137cd8303356dacc2a0abc24088a88263ef81f3a85b16ba4b950cc6bcb806f</code>, 46,002,244 bytes.</p>
     <p class="home-quiet" data-release-note-1><b>What changed in 2.0.1.</b> An update never leaves mining off. The Cards tab is back as its own page. The first-block card shows once per address. The window names included, executed, proven and finalised. The node serves its manifest and the genesis state stream to every miner. The Mac build is unsigned tonight.</p>
     <p class="home-quiet" data-release-note-2>2.0.0 is withdrawn: its miner could not read this network’s block templates; every 2.0.0 install updates itself to 2.0.1.</p>

site/ledger.html

@@ -273,7 +273,7 @@ blockquote{margin:10px 0;padding:10px 14px;border-left:3px solid var(--line-2);c
 <blockquote>GPU-friendly hashes always fall to chips. A specialist strips everything a GPU carries that the hash does not need.</blockquote>
 <div class="status"><span class="badge b-open">Open</span> <span class="did">8 October 2026): pass when the best supported cost and energy advantage of a re-optimised, programmable adversary sits inside the chosen competitiveness envelope, with its uncertainty published (D3).</span></div>
 <div class="pin"><b>Pin</b> The objective, property 1; D3 pass condition.</div>
-<details><summary>The answer as first written</summary><p>The target is contestable mining, not chip destruction: a manufacturer with a profitable product is not the failure; an exclusive, durable advantage large enough to displace the accessible GPU fleet is. Two measured results already bound the design. The connected-state reorganisation was killed as a class because rearranging the same operations moves neither side (D2). The mixed FP32 branch was killed because deterministic FP32 costs the cards 15 to 26 percent energy per hash against a 10 percent budget and widens the chip's edge to 3.0x to 3.2x (measured cards, modelled chip). Both stay as regression controls.</p></details>
+<details><summary>The answer as first written</summary><p>The target is contestable mining, not chip destruction: a manufacturer with a profitable product is not the failure; an exclusive, durable advantage large enough to displace the accessible GPU fleet is. Two measured results already bound the design. The connected-state reorganisation was killed as a class because rearranging the same operations moves neither side (D2). The mixed FP32 branch was killed because deterministic FP32 costs the cards 15 to 26 percent energy per hash against a 10 percent budget and widens the chip's edge (measured cards, modelled chip; the figure paired against class v4 is withdrawn). Both stay as regression controls.</p></details>
 </article>
 <article class="entry" id="R2" data-bucket="Open">
 <div class="head"><span class="id">R2</span><h3>Ordinary operators will be priced out</h3><span class="date">8 October 2026</span></div>

site/miner.html

@@ -297,7 +297,7 @@ pre b{color:var(--molten-text);font-weight:500}
           <div class="download-platform"><span class="osmark" data-os="hive" title="HiveOS"><svg viewBox="0 0 24 24" width="22" height="22" aria-hidden="true" focusable="false" fill="none" stroke="currentColor" stroke-width="1.7" stroke-linejoin="round" stroke-linecap="round"><path d="M12 2.6 20.2 7.3v9.4L12 21.4 3.8 16.7V7.3z"/><path d="M12 7.4 16 9.7v4.6L12 16.6 8 14.3V9.7z"/><path d="M12 7.4V2.6M16 9.7l4.2-2.4M16 14.3l4.2 2.4M12 16.6v4.8M8 14.3l-4.2 2.4M8 9.7 3.8 7.3"/></svg></span><div><h3>Ember for Linux and HiveOS</h3><p>a tarball for rigs and Hive flight sheets</p></div></div>
           <p class="download-note">For the rig people. One tarball, the miner and the node inside, the same signed manifest as the desktop apps.</p>
           <a data-dl="miner-hive" href="https://dl.igneum.network/public/igneum-miner-hive.tar.gz" class="btn primary"><svg class="icon" viewBox="0 0 24 24" aria-hidden="true"><path d="M12 3v12m-5-5 5 5 5-5M5 16v4h14v-4"/></svg>Download the tarball <span data-dl-meta="miner-hive" style="font-weight:400;opacity:.85">v2.0.1 · 52.7 MB</span></a>
-          <p class="fair" data-devnet-reset-note><b>The devnet is being reset this morning, 9 October.</b> A fresh install syncs once the reset completes. This line is re-read at 10:30 UK.</p>
+          <p class="fair" data-devnet-reset-note><b>Before you install: a fresh sync from genesis needs 64 GB of memory today.</b> 48 GB is the floor at this hour, and the number grows with the chain until the snapshot path ships (a fresh node reads about 30 GB of memory by DAA 44,000 and 35 to 40 GB at today’s tip, a gigabyte more every 17 minutes at one block a second; a 16 GB machine reaches DAA 7,000 to 8,000 about four minutes in and then swaps or is killed, a 32 GB machine about DAA 20,000). A smaller machine waits for the snapshot path, 2.0.3.1: its design is due by 12:30 today and the first served snapshot by 15:00; this note changes when it ships. The devnet was reset this morning, 9 October; a fresh install syncs from the reset chain.</p>
           <p class="fair" id="versions">Current build: Windows <span data-rm="versions.miner-windows.version">2.0.1</span>, macOS <span data-rm="versions.miner-mac.version">2.0.2</span>, HiveOS <span data-rm="versions.miner-hive.version">2.0.1</span>, the wallet <span data-rm="versions.wallet-mac.version">0.1.6</span>, <span data-rm="read_at_short">read 8 October 2026, 16:2x UTC</span>. The machine-readable list, with every file’s SHA-256, is <a href="/release.json">/release.json</a>.</p>
           <div class="hive-sheet" id="hive"><b>HiveOS Flight Sheet.</b> Miner: <b>Custom</b>. Installation URL: <code data-dl-url="miner-hive">https://dl.igneum.network/dl/public/igneum-hive-2.0.1.tar.gz</code>. Miner name <code>igneum</code>, wallet and worker <code>0x&lt;your 40-hex payout address&gt;.%WORKER_NAME%</code>. Hive itself is untested on our side: tell us what breaks.</div>
           <p class="sha">sha256 <span data-dl-sha="miner-hive">a16add6a9082747bc11d9bb9a69d193530f723dd2671fe1352ad59b1ae0a46ab</span></p>