constitution · epochs · watch · epoch 4

ev_54f8acb4767adee8dd1c2e6f245cbfe9a6347a4ec57de86e4c7c6f78540c6f09

links

• comparison_id: cmp_590a782531772267269aab0f22e7676d5d6bfe1f5b8934a045541f720cfdc104
• epoch 4

payload

{
  "comparison_id": "cmp_590a782531772267269aab0f22e7676d5d6bfe1f5b8934a045541f720cfdc104",
  "prompt": {
    "byte_length": 112461,
    "encoding": "blob",
    "sha256": "5c460e63e0edde1203025dde85e084ed53fc51fdabe8270f69e77e3f47b8bf85"
  },
  "ranking_run_id": "rank_db352e183a9b3e2cdaec085b51b3899c3b67306587a0561810a76295e52c5781",
  "side_a": {
    "commit_id": "c_d59e1908ed6f51f24dcbd75e9768738c77a610c2e35431ccb5c1073beded01df",
    "commit_ids": [
      "c_d59e1908ed6f51f24dcbd75e9768738c77a610c2e35431ccb5c1073beded01df"
    ],
    "contributor": "tommy-mor",
    "diff": {
      "byte_length": 3745,
      "encoding": "blob",
      "sha256": "b7b614d5e2639a2aa263e7996c542f672b3c6396f24ea45d3d400eebfffaf236"
    },
    "message": {
      "byte_length": 272,
      "data": "WzA3Y2MwZTFjXSBCaW5kIEhUVFAgYmVmb3JlIHJlcGxheWluZyB0aGUgbGVkZ2VyIHNvIGJvb3RzIHN0YXkgcmVhY2hhYmxlLgoKSW1wb3J0LXRpbWUgSnNvbmxTdG9yZSByZXBsYXkgb2YgYSA0NTBNQiBldmlkZW5jZSBsb2cgYmxvY2tlZCB1dmljb3JuIGZyb20KbGlzdGVuaW5nOyBkZWZlciBsb2FkIHRvIGEgc3RhcnR1cCB0YXNrIGFuZCBrZWVwIC9hcGkvaGVhbHRoIGFsd2F5cyBsaXZlLgoKQ28tYXV0aG9yZWQtYnk6IEN1cnNvciA8Y3Vyc29yYWdlbnRAY3Vyc29yLmNvbT4=",
      "encoding": "base64",
      "sha256": "9471728432454c22c84eb2fdf245be62a59479000ad54f0ca2794dcf7124e306",
      "text": "[07cc0e1c] Bind HTTP before replaying the ledger so boots stay reachable.\n\nImport-time JsonlStore replay of a 450MB evidence log blocked uvicorn from\nlistening; defer load to a startup task and keep /api/health always live.\n\nCo-authored-by: Cursor <cursoragent@cursor.com>"
    },
    "oid": "sha1:07cc0e1c5af5aecb3f88c4e225aa94809b4253e1"
  },
  "side_b": {
    "commit_id": "c_5d4f3299fa6dd70fc6719c17ac0a411f1a93bdda18866baca081f0db3eab4a6f",
    "commit_ids": [
      "c_5d4f3299fa6dd70fc6719c17ac0a411f1a93bdda18866baca081f0db3eab4a6f"
    ],
    "contributor": "tommy-mor",
    "diff": {
      "byte_length": 105575,
      "encoding": "blob",
      "sha256": "24805f3d370fbd6c1ebb1d37b1a1289acfd8c6c1ff34cbf5f34714b2089cf485"
    },
    "message": {
      "byte_length": 1930,
      "data": "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",
      "encoding": "base64",
      "sha256": "6893fdf92ec25e3f09b4140c25a7b0498222780f18c1d1af9bf2b70d8e1666a8",
      "text": "[0e81b5e1] Solve Rank Centrality exactly instead of capping power iteration.\n\n`compute_scores_from_edges` ran power iteration for at most 10 000 sweeps and\nreturned whatever vector it had reached, with no signal when the cap bound. On\na preference chain the spectral gap is Theta(1/n^2), so the cap binds past\nn ~= 1500 -- and worse, the L1 *step* tolerance is an absolute test, so on a\n1024-node chain with mixed ratios it declared success after 4325 sweeps while\nleaving 486 of 1024 items at the wrong rank. A 1024-node star fails outright:\nit needs 14 154 sweeps.\n\nNew `stationary` module owns every candidate solver -- power, Aitken-accelerated\npower, Gauss-Seidel/SOR, preconditioned BiCGSTAB, dense LU, dense GTH, and a\nsparse GTH state reduction with minimum-degree elimination -- behind one `solve`\nthat splits disconnected components, tries the sparse direct path first (trees\nand chains reduce in O(n) with zero fill), and falls back to Gauss-Seidel then\npower iteration on the dense graphs elimination cannot help with. Every answer\ncarries its own recomputed residual and a `converged` flag; nothing comes back\nclaiming success it did not earn.\n\nScores are also produced in log space. A chain of n items each preferred 2:1\nspans 2^(n-1), which leaves the f64 range at n ~= 1075, so `pi` genuinely cannot\norder a long chain no matter how it is computed. `Solution::log_pi` stays exact\nthere and is what `ranked_items_subset` now sorts on.\n\nEdges are sorted before aggregation, so the result no longer depends on\n`HashMap` iteration order. The legacy builder produced a different bit pattern\non 7 of 7 rebuilds of the same graph; the sort costs ~10% of chain construction.\n\nAdds `benches/stationary_solvers.rs` (~80s) and `examples/solver_probe.rs`\n(~30s), which reports time, residual, and rank displacement against closed-form\nanswers per topology. No new dependencies.\n\nCo-authored-by: Cursor <cursoragent@cursor.com>"
    },
    "oid": "sha1:0e81b5e1958ca486bffe87c800cc68e389c9a810"
  },
  "summary": "Comparing c_d59e1908ed6f51 (tommy-mor) with c_5d4f3299fa6dd7 (tommy-mor)"
}