The dataset covers filings whose claims or abstracts address software, source code or package builds described as reproducible, deterministic, hermetic, repeatable or bit-for-bit identical. In practice this spans compilation-environment control, build-artifact verification, tamper-resistant packaging and related methods used to confirm that a build output can be trusted and independently re-created. It also includes titles that use the specific phrases 'reproducible software build,' 'deterministic software build' and 'hermetic build.' The search string behind this landscape combines those title/abstract terms with claim-level language requiring both a build or compile step and a reproducibility qualifier.

Filing is not concentrated around one dominant holder: the leading assignee holds 11 of the 246 records in scope, and the top 5 assignees combined account for 20.3% of all records. The top 10 combined reach 32.9%, meaning even with the ten most active filers included, over two-thirds of the field's records sit outside that group. This points to a fragmented landscape where a large share of activity comes from a long tail of companies filing one or a small number of records each.

Filings peaked at 23 records in 2022 and fell to 7 by 2024, a -63% drop from the 19 filings recorded in 2021 — and 2024 is the most recent year this dataset treats as complete. Because patent publication typically lags the actual filing date by about 18 months, the lower counts showing in 2025 and 2026 are expected and do not yet indicate continued decline; those years are still filling in. The safest reading is that the 2021-2024 window shows a genuine pullback from the 2022 peak, without drawing further conclusions from the most recent two years.

The dominant classification is G06F (electric digital data processing), present on 74.4% of the 246 records, confirming that this is primarily a software and computing-platform problem rather than a networking or hardware one. H04L (digital information transmission) is the largest secondary class at 17.5%, followed by G06Q (business and commerce data processing) at 7.3% and G06N (AI-based computing) at 6.1%. Smaller classes such as B65G, H04M, H04Q and F24F each cover 3.3% or less of records and likely reflect niche applications of build-verification techniques rather than core claim territory.

The shallow concentration curve — a leader with only 11 of 246 records and a top 5 that reaches just 20.3% — suggests room for independently standing claims rather than a field locked up by one incumbent's portfolio. Technology-wise, classes adjacent to the dominant G06F cluster, such as G06N (AI-assisted build verification) at 6.1% of records, carry comparatively few filings relative to the core space. Jurisdictionally, the United States (122 of 246 records) and Europe (33) are the most contested; India at 19 records and the UK at 9 are lighter secondary tracks worth checking before assuming global coverage is already crowded.