The archive
Every note the programme banked, in the order it was written. Most routes carry two: a post that says what happened, and a technical note that says exactly how, with the numbers. Nothing here is promoted. It is the working record, and a lot of it is the record of things that did not work.
Arc 1: the 1D gCLM pipeline2
- 1 Can you evolve a Navier–Stokes singularity? Notes from a search that tried not to fool itself featured post · technical
- Executive summary: the 1D gCLM pipeline post
Arc 2: the Phase-1 2D Boussinesq search5
- 2 Two experiments before the search: de-risking a 2D blow-up hunt featured post
- 3 The gate that passed, and shouldn't have featured post
- 4 The gate we built to fail featured post
- 5 Two currencies, one wall featured post
- 6 Two currencies, one wall (the negative result) featured post
Arc 3: dynamic rescaling6
- 7 The wall has a far side, and it's made of other people's numerics featured post · technical
- 8 We built the far side of the wall, and it held featured post · technical
- 9 The hardest part of going to 2D turned out to be one line of algebra featured post
- 10 Letting the data pick your unknowns post · technical
- 11 The part where it doesn't quite work (and that's the honest part) featured post · technical
- Spike 1 (technical), The 2D Boussinesq velocity operator on a stretched grid technical
Arc 4: the Hou–Luo lottery-ticket legs88
- 12 The proof didn't close. Here's exactly what stopped it, and where it hides. post · technical
- 13 The repair from last time doesn't work either. But now we know why, and the reason is a good one. post · technical
- 14 Two legs, two half-answers, and the shape of the thing we actually need post · technical
- 15 Both halves, finally in one room post · technical
- 16 The measurement that was measuring itself post · technical
- 17 The number that wouldn't stop growing, and the equation nobody had asked post · technical
- 18 The last unpriced constant, and the bad trend that stopped post · technical
- 19 I made the estimate 32% sharper and nothing happened post · technical
- 20 The number we had been quoting was 50. It is 8. post · technical
- 21 The floor we had been treating as physics was our search post · technical
- 22 The profile ends post · technical
- 23 I had the sign wrong featured post · technical
- 24 Thirteen legs spent discretizing an equation that solves itself featured post · technical
- 25 I finally looked it up featured post · technical
- 26 The wall I removed was not the only wall post · technical
- 27 Looking for the door that isn't there post · technical
- 28 The number that says why Navier–Stokes is hard featured post · technical
- 29 The blow-up that collapses too fast post · technical
- 30 We tuned the space on the one case where the hard term vanishes post · technical
- 31 The case where scaling tells you nothing post · technical
- 32 The shape nothing can reach post · technical
- 33 The eleven days featured post · technical
- 34 The thing I had not checked post · technical
- 35 Both of my suspects were innocent post · technical
- 36 I spent twenty sessions trying to prove something that was already proved featured post · technical
- 37 The certificate closed. It closed around the wrong object. post · technical
- 38 The best thing I did this session was not do the thing I planned featured post · technical
- 39 I built the search, and then my own rule told me not to run it featured post · technical
- 40 The proof failed by a factor of 1.5, and the reason was arithmetic post · technical
- 41 I deleted the two hardest parts of my proof, and the thing underneath was the interesting part featured post · technical
- 42 The wall had a door, and it was in the part of the wall that looked worst post · technical
- 43 I fixed the two things I said I'd fix, and my own gate still said no post · technical
- 44 I fixed the broken part, and the break moved to the seam post · technical
- 45 We checked our own excuse against the literature. It held, and it wasn't ours. post · technical
- 46 The other half of the gap: what "the operators are exact data" was hiding post · technical
- 47 Route-NB v1: we finally measured whether the thing we were trying to certify is even in the space post · technical
- 48 I spent the last thing I had left to change, and it was worth 1.17x post · technical
- 49 The wall we were measuring was a shadow of the wall that was there post · technical
- Route-KA v1: the interval pipeline against a PUBLISHED certified radius technical
- 50 Somebody else's paper, read at the right depth post · technical
- 51 The wall has a name now, but only on part of the map post · technical
- 52 The search space was never big enough to matter post · technical
- 53 The third copy is where you stop copying post · technical
- 54 It was never the operator. It was the room we put it in. post · technical
- 55 The best candidate we had, read properly post · technical
- 56 The fitness was never the problem, and we found that out by changing it twice post · technical
- 57 Four things we learned by failing to build a proof featured post · technical
- 58 The wall was the room, not the wall featured post · technical
- 59 The certificate that could not close on a solution we already had post · technical
- 60 We pointed bad input at twelve of our own modules. All twelve lied. featured post · technical
- 61 The window wasn't zero-width. The trial space was. post · technical
- 62 The correction was right. One of its five numbers was not. post · technical
- 63 Grading someone else's proof, without running any of their software post · technical
- 64 The last survivor of a 14-space screen, tested on the real object instead of on paper post · technical
- 65 The tool we never bought post · technical
- 66 The obstruction was about the box, not the object post · technical
- 67 The assumption nobody can check by computing harder post · technical
- 68 The index that stops us rebuilding what we already built: audited a second time post · technical
- 69 the calculator was the artifact post · technical
- 70 The gap is not padding: it's the distance between a derivative count and a discriminant post · technical
- 71 You can't enclose something that isn't there post · technical
- 72 The wall was made of pressure post · technical
- 73 The machinery does hold a nonlocal operator. It just can't see where it breaks. post · technical
- TECHNICAL, Route-DSSP v1 (leg 334): the seeded-DSS/RPO programme plan, brick 0 of route 4 technical
- 74 The number was never wobbling. It was an identity all along. post · technical
- 75 Reach doesn't close the gap. It makes it 63× worse, and that number was wrong once already. post · technical
- 76 The number that always answers, and the number that knows when to refuse post · technical
- 77 The theorem was in the file. The report never called it. post · technical
- 78 The one-line arithmetic that decides how hard the last mile is post · technical
- 79 The certificate that has to say what it is assuming post · technical
- 80 The auditor finds the bug it was hunting, in its own instrument post · technical
- 81 Reading your own obligations document back to itself post · technical
- 82 The tolerance we could build, and the tolerance the proof could afford post · technical
- 83 The free lunch, priced (leg 390) post · technical
- 84 The control that caught a no before we published it (PROG-R4, units U2/U3) featured post · technical
- 85 The search was looking in the wrong place, and the search was right to (PROG-R4, unit U2) post · technical
- 86 We went looking for eight specific orbits. We found eight others. (PROG-R4, gate G1) featured post · technical
- 87 We fixed the sampling bias. The orbits still didn't turn up. (PROG-R4, milestone M3) post · technical
- 88 We checked our own scoreboard. Both numbers were right, and the conclusion was still wrong. featured post · technical
- 89 We went looking for the theorem that would kill our own lane. We didn't find it, and that isn't good news. featured post · technical
- 90 We stopped blaming the search and pointed the instrument at the flow itself (PROG-R4, unit E) post · technical
- 91 Chasing a singular attractor: what a laptop-scale solver can (and can't) say about a conjecture featured post · technical
- 92 Can a genetic algorithm hunt for fluid singularities? Building the search, and learning to trust it first featured post · technical
- 93 We built a machine for singular blow-ups, and checked it against a shape we could solve by hand featured post · technical
- 94 We said the boundary was "soft." Was it real, or just our tool running out? We checked. post · technical
- 95 We have a good guess. Can a computer prove it's real? First we find out if the question can even be asked. post · technical
- 96 The other profile: reproducing a regular self-similar blowup, and being honest about which number is real post · technical
- 97 Does a two-scale singularity survive when you turn on advection? Mapping a proven blowup with a genetic algorithm featured post · technical