Public-Key Subset Preaggregation for BLS Certificates: Differential Checks and Historical Benchmark Limits

This report examines an uncommitted Agave prototype that precomputes public-key subset sums for groups of four validators, then reuses them across certificate signer bitmaps. Its specific historical user/AI conception remains unresolved; Jarett Dunn is the author of this analysis, not an established inventor of the prototype. A fresh isolated rebuild passed twenty existing unit tests, including four PreparedRankMap differential tests, using only ephemeral synthetic BLS keys and signatures. Sixteen historical benchmark comparisons match their archived logs and arithmetic, but were not rerun or authenticated against the original hardware and source bytes. The recorded dense-case improvements coexist with sparse regressions. The implementation retains upstream signature and stake checks, while production epoch, fork and reward integration remains unverified. These results establish bounded artifact and test observations, not cryptographic security, consensus safety or whole-validator savings. Human agency, financial inclusion and evidence-based accountability shape the evaluation through honest resource claims, access questions and traceable provenance. No measured social outcome or exclusive invention is established. Jarett Dunn. Technical report, version 1.0. Preprint, not peer reviewed. Codex assisted source recovery, analysis, local tests and writing. The manuscript and new analysis are CC BY 4.0; software retains its original licenses and notices. Prototype authorship remains unresolved.