| # Compute and Timeline | |
| Status: **PAUSED**, not complete and not abandoned. | |
| - Public-corpus hydration began: 2026-08-11T22:48:50+00:00 | |
| - First verified research package: 2026-08-12T04:50:16+00:00 | |
| - Pause boundary (last completed package before the operator paused the campaign): 2026-08-12T17:00:21+00:00 | |
| - Hydration-to-pause elapsed time: **18 h 11 min 31 s** | |
| - Verified-package generation span: **12 h 10 min 05 s** | |
| - Verified research packages: **49** | |
| - Distinct selected papers covered: **34 of 48** | |
| - Selected papers remaining at pause: **14** | |
| - Persistent verified-result records present at pause: **3,077** | |
| - Aggregate release-gate issues after normalizing one valid nonstandard status schema: **0** | |
| ## Rough token cost | |
| The current campaign is estimated at **approximately 3-8 million model tokens**. Order-of-magnitude estimate: 49 research packages at roughly 40k-130k model tokens per package, plus corpus hydration, planning, verification, synthesis, and publication-stage overhead. No billing-grade aggregate token counter was recorded. This range excludes the separate prior recursive-self-improvement campaign. | |
| ## Hardware counterfactual | |
| This was laptop-bound work. A faster laptop—especially more CPU throughput, faster storage, more memory headroom, and greater safe parallelism—would have permitted more research/verification cycles in the same wall time. That is a throughput counterfactual, not a claim that hardware alone guarantees stronger theorems. | |
| ## Prior recursive-self-improvement cycle | |
| This campaign was deliberately run **after** a separate, operator-reported recursive-self-improvement cycle lasting roughly 29 hours. That prior cycle changed workflows, retrieval, verification, orchestration, and persistent capability memory; it did not change model weights. The underlying operator evidence is not part of this research package. | |