Nothing is solved. This is the record of the sub-questions the lab actually closed between November 20 and 21, 2026, each with its round, its receipt IDs, and its data source.
Nothing is solved. None of the six Millennium Problems is solved, and nothing in this document claims otherwise. What follows is the record of the sub-questions the lab actually closed between 2026-11-20 and 2026-11-21, each with its round, its receipt IDs, and its data source, so that a stranger can check every line. The lab is research, not a solution, and the researcher takes no credit for the underlying mathematics: others did all of that work; this lab threw a little LARP on top.
Lab state at time of writing: lab v6.0, 66 receipts, chain valid, head bdebd1e7d93dd108bffe0ba01d499e603db81a8c17027bb0aad36a70f73dacdd. Receipt ledger: ~/workspace/millennium-lab/data/lab.db (events table), mirrored in receipts.json. Timestamps in this file are UTC as recorded in the ledger.
0 -> Z -> O_X -> O_X* -> 0 on the opens site, with the connecting homomorphism c1, proved in Lean 4.13.0 + mathlib v4.13.0, 18 new theorems, 0 sorries.rcpt_0057 (formalization). Chain verification for the round: rcpt_0059.~/workspace/millennium-lab/data/hodge_r5.json (theorems_proved, 18 entries, all sorry: false); Lean source ~/workspace/millennium-lab/lean/Hodge/LongExactSequence.lean; lake build exit 0, 4,584 / 4,584 targets.O_X on the opens site builds and its basic properties prove in Lean. This overturns the round 2 assessment that it was not tractable.rcpt_0047 (formalization, round 3). The round 2 assessment it overturns is rcpt_0041.~/workspace/millennium-lab/data/hodge_r3.json; Lean source ~/workspace/millennium-lab/lean/Hodge/SheafOX.lean; lake build exit 0, 3,842 / 3,842 targets.rcpt_0051 (formalization). Chain verification for the round: rcpt_0053.~/workspace/millennium-lab/data/hodge_r4.json. Note the counting subtlety, verified in the 2026-11-21 second-pass audit: theorems_proved in that file lists 18 entries, which is 14 new Dolbeault theorems plus 4 re-verified theorems from the previous agent's work; theorems_checked is 23. The "14 new" claim is correct, and the poster states both numbers.~/workspace/millennium-lab/lean/Hodge/Dolbeault.lean; Dolbeault-target build exit 0, 1,497 / 1,497 targets.s and every timeline, exactly one of leaf(s) and NOT(defeater s) is SATISFIED. 8,876 checks, 0 violations.rcpt_0039 (round 2 audit, greedy result); rcpt_0046 (round 3 audit, dichotomy). Chain verification: rcpt_0043 (round 2), rcpt_0053 (rounds 3-4).~/workspace/millennium-lab/data/p_vs_np_r2.json (greedy), ~/workspace/millennium-lab/data/p_vs_np_r3.json (dichotomy: 14 timelines x 317 signals x 2 floor settings, 2.02 s). Brute force agrees: full fragment 47/47, defeater-free 44/44, extended 44/44.p^r sharply, a textbook phase transition. With defeaters present, the transition collapses: at every tested density, 200 of 200 trials find a witness. 1,400 of 1,400 trials SATISFIED across 7 densities (p = 0.0, 0.1, 0.25, 0.5, 0.75, 0.9, 1.0).rcpt_0050 (audit). Chain verification for the round: rcpt_0053.~/workspace/millennium-lab/data/p_vs_np_r4.json (existence table, 1,400 trials; witness kinds NOT(defeater) vs plain loose leaf); phase table in ~/workspace/millennium-lab/data/p_vs_np.json (measured values: p=0.5 gives 0.52 at r=1, 0.235 at r=2, 0.115 at r=3, 0.02 at r=5, 0.01 at r=8, 0.0 at r=12 and r=20; p=0.25 gives 0.335 at r=1, 0.085 at r=2, 0.01 at r=3, 0.0 from r=5 on; p=0.75 gives 0.74 at r=1, 0.47 at r=3, 0.255 at r=5, 0.1 at r=8, 0.035 at r=12, 0.0 at r=20; 200 trials per cell).rcpt_0056 (audit). Chain verification for the round: rcpt_0059.~/workspace/millennium-lab/data/p_vs_np_r5.json (reduction_check per_op; CONTRADICTION_FREE excepted, see below).lfunzeros at real precision 30, all with max Re deviation 0.0 from Re(s) = 1/2. Counts per round: 1,000 / 2,000 / 3,000 / 4,000 (cumulative, first N zeros re-used each round). First zero t = 14.13472514173469379045725. Last zero t: 1,419.42 (1,000) / 2,515.29 (2,000) / 3,533.33 (3,000) / 4,506.31 (4,000). Von Mangoldt count check agrees every round (expected 999.42 vs 1,000; 1,999.38 vs 2,000; 2,999.37 vs 3,000; 3,999.53 vs 4,000).rcpt_0001 (round 1, first 1,000 zeros); rcpt_0042 (round 2 calibration, carries both bsd and riemann payloads); rcpt_0052 (round 4 calibration, same); rcpt_0058 (round 5 calibration, same). Chain verification: rcpt_0030 (round 1), rcpt_0043 (round 2), rcpt_0053 (rounds 3-4), rcpt_0059 (round 5).~/workspace/millennium-lab/data/riemann_zeros.json, riemann_zeros_r2.json, riemann_zeros_r4.json, riemann_zeros_r5.json. The first ten t-values are exact strings in riemann_zeros.json (14.13472514173469379045725 through 49.77383247767230218191678).rcpt_0002 through rcpt_0029 (round 1, one per curve, 28 curves); rcpt_0042 (round 2, 57 curves); rcpt_0052 (round 4, 93 curves, 92 rank 0 and 1 rank 1); rcpt_0058 (round 5, 130 curves, 128 rank 0 and 2 rank 1). Chain verification: rcpt_0030, rcpt_0043, rcpt_0053, rcpt_0059.~/workspace/millennium-lab/data/bsd_curves.json, bsd_curves_r2.json, bsd_curves_r4.json, bsd_curves_r5.json. The JSON field is named L_value but for rank-1 curves it records L'(E,1); the lab page and the BSD poster both state the derivative reading.rcpt_0060 (round 6 computation, 64^3 nu 0.000125, T = 10 and T = 20); rcpt_0063 (round 7 computation, full 96^3). Chain verification: rcpt_0062 (round 6), rcpt_0066 (round 7).~/workspace/millennium-lab/data/navier_stokes_hunt_r6.json (SHA-256 b6a88cb2dbcd02e89300814dbf3f7d8f1d3b83d8b05faaabf00e7adb0e420254), navier_stokes_hunt_r7.json (SHA-256 5ac93078cac19b6f0986b7148bb845fee3d83ef3bc174b40211a15ec984e6490). Dual runs produced identical field hashes (round 6 Kida-Pelz T = 20 hash 35e01ce6..., round 7 Kida-Pelz 96^3 hash 6445c2c1...).rcpt_0060 (round 6, cost measurement); rcpt_0063 (round 7, full run). The round 3 rejection it overturns is rcpt_0045.~/workspace/millennium-lab/data/navier_stokes_hunt_r6.json, navier_stokes_hunt_r7.json.Everything in this section is still open. Nothing here was resolved, and nothing here is claimed to be close.
H^q(X, O_X) ≃ H^{0,q}(X). mathlib v4.13.0 has no differential forms on manifolds, so the Dolbeault groups cannot be stated. Blocked on the library, not on the idea.Things the lab tried, believed, or was told, that the lab's own measurements killed. Each one is a result: knowing a channel does not work is worth recording.
rcpt_0061 (round 6, second seed); rcpt_0064 (round 7, third seed and estimator test).~/workspace/millennium-lab/data/yang_mills_r6.json (SHA-256 41ffee8393f21b33bc5e1b43cd29fe5d0fa803ae3e4190033323007339ad5cf2), yang_mills_r7.json (SHA-256 9abc2cfe61a98b2810988dd1928d049544388b7d077c665f48cf799feaf194c1). Three-seed mean fit: intercept 2.120 +/- 0.526, chi2/dof 55.35/3. No extrapolation is attempted from it.p^r. The full fragment with defeaters does not. The transition is a property of the defeater-free fragment, not of the problem as posed.rcpt_0050.~/workspace/millennium-lab/data/p_vs_np_r4.json, p_vs_np.json.rcpt_0045 (rejection); rcpt_0060, rcpt_0063 (overturn).rcpt_0041 (assessment); rcpt_0047 (overturn).rcpt_0061.~/workspace/millennium-lab/data/yang_mills_r6.json.rcpt_0060.~/workspace/millennium-lab/data/navier_stokes_hunt_r6.json.A full audit of every component of the program from the last 4-5 days, run before this document was written. Everything was read from the files, not from memory.
Checked and verified:
- data/lab.db: 66 events, head bdebd1e7, chain valid. Kinds: 47 COMPUTATION, 7 VERIFICATION, 7 AUDIT, 5 FORMALIZATION. Receipt ranges: rcpt_0001-0037 rounds 1-2 (to 2026-11-20 17:43 UTC), rcpt_0038-0053 rounds 3-4 (to 2026-11-21 00:54 UTC), rcpt_0054-0059 round 5 (2026-11-21 05:07 UTC), rcpt_0060-0062 round 6 (12:49 UTC), rcpt_0063-0066 round 7 (15:04 UTC). receipts.json: 66 entries, matching.
- BSD counts per round (28 / 57 / 93 / 130), rank split (128 rank 0, 2 rank 1 at conductor 50), L' values for 37a1 and 43a1, all verified against bsd_curves*.json.
- Riemann counts per round (1,000 / 2,000 / 3,000 / 4,000), t_first 14.13472514173469379045725, t_last values, max Re deviation 0.0 every round, all verified against riemann_zeros*.json. The round calibration receipts (rcpt_0042, rcpt_0052, rcpt_0058) carry both bsd and riemann payloads.
- P vs NP: dichotomy 8,876 checks 0 violations; brute force 47/47, 44/44, 44/44; existence 1,400/1,400; phase table values verified against the evidence string in p_vs_np.json; greedy extension 5 of 6 with the CONTRADICTION_FREE exception (10 zero-arg, 2,664 one-arg violations) verified against p_vs_np_r5.json.
- Hodge: new theorems per round 4 / 9 / 13 / 14 / 18, 0 sorries every round; builds 3,060/3,060 (R2), 3,842/3,842 (R3), 1,497/1,497 (R4 Dolbeault target), 4,584/4,584 (R5); axioms [propext, Classical.choice, Quot.sound], no sorryAx. Lean sources present: Basic, ChernClass, Dolbeault, Exponential, LongExactSequence, SheafOX.lean.
- Navier-Stokes ladder 18.86 / 40.33 / 64.46 / 88.24 (64^3), 96^3 145.79 / 138.27, ratios 1.65 / 1.58; T = 20 transience for both Kida-Pelz and Taylor-Green; dual-run field hashes identical.
- Yang-Mills three-seed torelon values and tau spreads, verified against yang_mills_r7.json.
- Lab page v6.0 (95,128 bytes): all key numbers present and matching the data files.
- Atlas ~/workspace/six-unsolved/index.html (40,750 bytes, MD5 1a6d982c8a1e8a64f59afa369539b1aa).
- Four problem pages plus problems.html in ~/workspace/millennium-share/: key numbers spot-checked, all nav links resolve, zero em dashes.
- Poster renders in ~/workspace/millennium-lab/poster/: 10 field renders (rounds 1-7) plus the round 2 poster, all 1080x1350, computed from the lab's own solvers.
- Shared artifact (slug millennium-problems): live and public, verified via artifact status.
Found and fixed during this pass:
- The four problems BSD, Riemann, P vs NP, and Hodge had charts inside HTML pages but no standalone poster images. Built 2026-11-21 (~/workspace/millennium-lab/poster/build_four_posters.py, matplotlib, data read from the JSONs, bone/paper + ink + brass): millennium-bsd-poster.png (169,782 bytes), millennium-riemann-poster.png (185,042 bytes), millennium-p-vs-np-poster.png (223,513 bytes), millennium-hodge-poster.png (210,227 bytes). All four read at native size and verified number by number against their data files. Copies in ~/workspace/millennium-share/renders/ (MD5-identical). Companion meta JSONs alongside them in poster/ record every plotted number and its source file.
- During visual verification, six layout defects were found and fixed (rank-1 labels colliding with bar totals on the BSD poster; the 43a1 annotation overlapping a histogram bar; the Riemann von Mangoldt box covering bar labels, then running off the figure edge when moved; the P vs NP phase-table parser dropping decimal values because its regex stopped at the first period; the P vs NP findings box overstating the greedy extension as 6 of 6 instead of 5 of 6; the P vs NP figure note colliding with the next panel title). The phase-table parse and the 5-of-6 correction are the two that touched numbers; both are now verified against the source files.
- The R4 Hodge counting subtlety (18 entries listed, 14 new + 4 re-verified) is documented above under item 3.
Stale or unverified items, still open:
- Lab page v6.0 carries the line "deployed but not publicly readable until it is flipped back." The page is not deployed anywhere public: the hub was deleted 2026-11-21 and the demos repo is private. The live copies are the shared artifact and the local files. Fix on the next lab page update.
- The four new posters are not yet in the shared artifact (it was built before them). They live in poster/ and millennium-share/renders/; adding them to the artifact is a rebuild, not yet ordered.
- No round 8 is ordered. Round 8 appends, when ordered, go through ~/workspace/millennium-lab/compute/receipts_append_TEMPLATE.py (MINCAP-stamped payloads, EXPECTED_COUNT_BEFORE = 66). Never run compute/receipts.py on the live ledger: it deletes the ledger and rebuilds rounds 1-2 only.
Written 2026-11-21, second pass over rounds 1-7. Every number above traces to a data file or a receipt in ~/workspace/millennium-lab/data/. Where a number could not be traced, it is not in this document.