REVIEW NOTE: ONE CONCEPT FROM A REAL SCAN A filled-in example for https://obviouslynot.ai/reviewing-concept-scanner-results The full walkthrough of this scan: https://obviouslynot.ai/how-concept-scanner-works Concept: Dual-cache with eviction prioritization Repository: https://github.com/golang/groupcache (Apache-2.0) Commit read: 2c02b8208cf8c02a3e358cb1d9b60950647543fc (no uncommitted changes) Location: groupcache.go, lines 346 to 469; the rule at lines 352 to 368 Scan: Concept Scanner v4.1.0, Balanced profile (qwen3-coder:30b, graded by gpt-oss:20b), local Ollama, 19 November 2026 Decision: Approved PROBLEM A server keeps two caches: a main cache for the keys it owns, and a smaller hot cache of popular keys copied from other servers. When their total is over the memory limit, one of them has to give up an entry. MECHANISM (corrected) When the total is over the limit, evict the oldest entry from the hot cache if it holds more than an eighth as much as the main cache, and from the main cache otherwise. Repeat until the total fits. CORRECTION MADE AGAINST THE SOURCE The scan wrote: "evicting from the larger cache when total usage exceeds limits." The code does not pick the larger cache. With a 64 MB main cache and a 6 MB hot cache, the main (larger) cache gives up an entry; at 10 MB of hot cache, the hot (smaller) one does. WHAT DIFFERS FROM THE OBVIOUS APPROACH The scan listed: one LRU cache with a fixed policy; separate caches each evicting on its own. This code uses one rule across both caches, which holds the hot cache to about an eighth of the main cache as load shifts. EVIDENCE - Cited line found in the file: "if hotBytes > mainBytes/8 {" (line 364). - Tests paired by the scan: 8, among them TestCacheEviction (groupcache_test.go:189). Read, not run. Run them before relying on them. - The comment above the rule: "this is good-enough-for-now logic. It should be something based on measurements and/or respecting the costs of different resources." FIGURE Five components (main cache, hot cache, cache bytes limit, populate cache, victim pointer). The scan's drawing missed the rule; we added two arrows: "chosen when hot bytes exceed 1/8 of main" and "chosen otherwise". PEOPLE AND DATES git blame: 41 lines by Brad Fitzpatrick, all in the repository's first commit, 23 July 2013. Code authors are not necessarily the inventors; who conceived the rule is a question for the people involved and a practitioner. Public since: 23 July 2013 (open-source repository). OPEN QUESTIONS For the engineer: why an eighth? Was another ratio measured? What happens to hit rates if the hot cache has no limit? For the practitioner: the code has been public since 2013. What does that mean for anything built on it? OTHER DECISIONS IN THIS SCAN Rejected as duplicate: a second concept describing the same eviction rule. Rejected as routine: choosing an HTTP transport per request. A scan describes engineering. Whether a mechanism is new, who the inventors are and whether to file are questions for a patent practitioner.