Across the 178 records in this dataset, filings are ranked among 43 assignees, and the leader holds 36 records — more than three times the fifth-place total of 10. The top 5 assignees combined account for 110 records, or 61.8% of all records in scope, and the top 10 reach 80.3%. That concentration means a small group of platform, JVM and chipmaker vendors set the terms for most enforceable claims in dynamic and tiered compilation.

Filing peaked at 12 records in 2019 and has since eased; the most recent complete comparison, 2021 to 2024, shows a -25% move from 4 records to 3. Years after 2024 look lower still, but that is largely an artefact of the roughly 18-month lag between filing and publication, not a real collapse in activity. The honest read is a mature technique with modest, not dramatic, decline in new filing.

98.3% of the 178 records in scope carry a G06F classification, meaning they are filed as core electric digital data processing rather than as a specialised or cross-domain technology. Small pockets of overlap exist with image data processing (G06T) and wireless/network transmission (H04B, H04L, H04W), each covering just 1.1% of records. That pattern suggests the field is defended primarily as general-purpose computing infrastructure, with only narrow forays into adjacent domains.

US20170052804A1, filed by International Business Machines Corporation and published 2017-02-23, claims a method for controlling the priority of dynamic compilation threads based on a monitored metric such as compilation-queue length or utilization rate. When that metric crosses a threshold, the returning thread's priority is lowered, for instance by making it sleep before re-entering the thread pool. It is a scheduling-layer control on the compilation process itself, not a compilation technique, so it primarily constrains how competitors manage compiler thread priority rather than how they compile code.

The technology composition data shows only 1.1% of the 178 records touching each of G06T, H04B, H04L and H04W, which points to under-claimed intersections between core JIT mechanics — inline caching, on-stack replacement, compilation thread scheduling — and imaging or wireless transmission contexts. Co-filing is also rare, with only 10 co-assignee pairs in the whole corpus, suggesting little collaborative patenting to build on. A first claim in one of these narrow branches would need to tie a specific compilation mechanism to a specific transmission or imaging use case rather than restate the general technique.