{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreibv5faog6zqmfewx43ck4pzwohe6vn6xtxc5zs53bc44g4ei2sou4",
    "uri": "at://did:plc:pgryn3ephfd2xgft23qokfzt/app.bsky.feed.post/3mo7vmhqe4ts2"
  },
  "path": "/t/shannon-prime-lattice/176466#post_14",
  "publishedAt": "2026-06-14T03:52:52.000Z",
  "site": "https://discuss.huggingface.co",
  "tags": [
    "04 — The Oracle & the Teacher",
    "05 — The Probe Suite",
    "06 — Computing on the Zip File",
    "07 — The Auditable Latent Crossbar",
    "08 — O(1) KV: a context-decoupled cache",
    "09 — KAIROS: a resident 12B daemon",
    "10 — Receipts or it didn’t happen",
    "@8"
  ],
  "textContent": "## **The paper series**\n\nA staggered set of short, independently citable, receipts-first papers — each carries its own one-command reproduction.\n\n  * **01 — Two-ring memory** — query-directed recall + byte-addressable KV offload (the needle-off-NVMe result above).\n  * **02 — The reducing loader** — output-preserving transcode + zero-copy load (the ~50%-smaller, bit-faithful result).\n  * **03 — Frobenius calibration-free quantization** _(staged)._\n  * **04 — The Oracle & the Teacher** _(written)_ — oracle-grounded backend verification: KL 2.7e-10 port, teacher-forced decode — plus the case study where a hand-written oracle measured gemma-4’s true PPL at **4.68** and convicted the GGUF ecosystem (192–506) while exonerating llama.cpp’s forward.\n  * **05 — The Probe Suite** _(written)_ — bisection, isolation and benchmark hygiene **as one set** — from the 12.65× phantom and the 0/256 K-quant bug to ecosystem-scale forensics and simulate-before-build (artifact matched the simulator to four decimals).\n  * **06 — Computing on the Zip File** _(complete, citable)_ — the dp4a bandwidth ladder (f32 1× → int8 ~3.8× → Q4 ~7.06×), the OK_Q4B block-scaled kernel, the sovereign quantization pipeline, and the gated headline: **26.1 tok/s at PPL 5.12 on an RTX 2060 12GB**.\n  * **07 — The Auditable Latent Crossbar** _(staged draft)_ — a frozen 12B steered by direct KV-cache transplant, no tokens: **15/15 incorporation + 15/15 selectivity** , self-transplant null 7/7 bit-identical, gold-instrument coherence (X-R1).\n  * **08 — O(1) KV: a context-decoupled cache** _(staged draft)_ — a learned **512×32 LSH router** decouples the KV cache O(1) from context: **+0.47% PPL @8×** (oracle −0.08%), **O(1) VRAM (8k↔16k flat ~50 MiB)** , **NIAH needle survives 10/50/90%** (frozen-router control misses) (X-R2).\n  * **09 — KAIROS: a resident 12B daemon** _(staged draft — release gated on the in-flight soak)_ — disciplined silence + **O(1) bit-exact rewind** : 24-tick crucible perfect, rewind byte-identical across 48 layers, metal **0.0073** vs prefix-grow **0.924 s/action** (KAIROS-01/02). The **≥24 h endurance soak is running now** — no verdict from a mid-run log; the paper ships on the receipt.\n  * **10 — Receipts or it didn’t happen** _(staged draft)_ — bit-exact-or-bounded as the contribution: the four documented self-corrections that prove the gates discriminate.\n\n",
  "title": "Shannon Prime Lattice"
}