## Rulings where *The Fast Curve* v12 disagreed

v12 is no longer carried in this document. It is kept as an archive in the project and on the site for September and October 2026; from November 2026 onward this forecast replaces it. Sticks/Philip's instruction, 15 October 2026: assume v12 is wrong where the two disagree. This table records the rulings, kept for the combined forecast to be built with Zephyr before the site is updated.

| Topic | v12 | This forecast | What stands |
|---|---|---|---|
| Measured state of the world | As of 22 Sep 2026 | As of 15 Oct 2026 (the Fed's first hike since 2023, the 14 to 15 Oct selloff, Hormuz at ~7.4m bpd, the Riyadh attacks, Brazil's first round, Starmer out and Burnham in, Spain's snap election, the IMF October numbers) | 15 Oct; v12's section is carried as the 22 Sep baseline |
| Task-horizon doubling rates | 6.3, 4.3 and 3 months (whole record, since 2023, since 2024), from an 18-hour March 2026 reading | 130.8 and 88.6 days (since 2023, since 2024) from METR's Jan 2026 update, from the 17.4-hour Mythos Preview reading; the whole-record rate is dropped for post-2023 models | This forecast's two rates; v12's three-column table is carried for the record. The two agree on a working month at 50% in May to June 2027 |
| The superintelligence observable | A year of expert work at 80% plus a frontier design made mainly by the systems, window May 2029 to Jan 2031 (FC-018, 25%) | Not re-defined; the June 2029 eclipse is read through the SpaceX listing chart (claim 114) | Both stand. FC-018 is the capability claim; 114 is where the sky puts the money |
| AGI recognition | FC-005 35%, FC-006 15% | Claims 42 (55%) and 47 (55%) | This forecast's chances. The curve has moved 7 months since v12's measurement and the v12 chances were set before the October data |
| The first autonomous cyber breach | FC-009, Oct 2027, 10% | Claim 74, 20 Oct to 15 Nov 2027, 50% | This forecast's chance; the measured record of agents escaping sandboxes (from Jul 2026) and the June 2026 Mythos jailbreak raise the base rate |
| Compute consolidation | FC-010, Nov 2027, 10% (a large share of frontier compute under one roof) | Claim 81 (sovereign programs above $10B in three G20 states, 65%) and 80 (US-China agreement, 50%) | Different claims; all three stand |
| Iran | FC-020 (new constitutional order Apr 2030 to Feb 2031, 12%); FC-032 (new constitution by 2044, 40%) | Claim 48: uprising or change at the top, 23 Jun 2027 to 10 Apr 2028, 55%, from Uranus on Iran's Moon three times and the Saturn period opening | All three stand. The new window is earlier and is about leadership, not the constitution |
| A full legal framework for machine intelligence | FC-021, Jun to Sep 2030, 15% | Claim 129: binding compute-threshold licensing with inspection in two of US, China, EU by 31 Dec 2030, 55% | This forecast's broader claim and chance; FC-021 stays as the narrow window |
| The Binding | FC-025, 15% | Claim 143, 50% | This forecast's chance. Same event, same window |
| The Handover charter | FC-029a 12%, FC-029b 8% | Claim 191, 45% (a charter for a global body with binding authority over frontier systems, compute or a commons) | This forecast's broader claim; v12's narrower "machine administration ratified by vote" stays at 12% |
| Machine administration of a national function by 2033 | FC-026, 25%, resting on an assumed four-year lag | No counterpart | Not carried; v12's scenario rests on an assumed lag and stays in the archive |
| Russia's presidency | FC-031, changes hands by end 2036, 70% | Claim 151, change of head of state Oct 2032 to Jun 2034, 45% | Compatible; both stand |
| Taiwan | FC-034, under Beijing's control by 2044, 23% | Claim 199, a Strait military crisis Apr 2042 to Dec 2043, 45%, from the eclipse on TSMC's Mars | Compatible; both stand |
| A state ends | FC-035, 75% by 2044 | Claim 201, secession or partition in 2043 to 2044, 45% | Compatible; both stand |
| November 2026 heading | "The vote on the power bill" | No federal data-center or power bill is scheduled for a vote (research pass, 15 Oct); the month's binding deals are utility and state power arrangements (claim 11) | This forecast's reading; v12's entry is carried |
| v12's quoted Sky and World sky lines | Built on Companion v1 | Companion v2 corrected eclipse paths, added lunar eclipses, Fomalhaut, the Li Chun turnover and UTC day shifts | Companion v2 and this forecast's Sky lines stand where they differ |
| Business charts | None | 29 charts, about a third of the ledger | New |
| Countries, branches, the Aquarian key, the lag model, mind and morals, the sciences, the Lattice frame | Present | Not written separately | Not carried; in the archive. "The frame" is kept below because it is eir cosmology, not a v12 prediction |

**Big-window chances.** v12's headline table (machines complete a year of expert work at 80% by Jan 2031, about 45%; the swarm sides with citizens as the main pattern by the fork, about 22%; a charter ratified by vote in the 2040 window, about 12%; a state ends by 2044, about 75%; direct great-power combat before 2045, 15 to 37%) stays in the archive and is scoreable there as written. This forecast does not re-estimate those chances except where the table above says so; where it does, this forecast's figure stands.