All posts · page 3

Deep-dive analysis of luau-lang/luau: 534,195 lines across 1,533 files, deferred constraint resolution, optional native codegen, 16,270 test-assertion matching lines, and 1,822 invariant assertions.

Analysis of the federated ActivityPub engine developed by Eugen Rochko and the Mastodon community with prod-code: 189,749 lines across 3,278 files, ActivityPub verification, and 8,191 RSpec expectation matches.

Deep-dive analysis of Neovim's modern Lua architecture with prod-code: 448,172 lines across 854 files, asynchronous LSP client, incremental document synchronization, and reproducible source-pattern counts.

Historical Nim compiler analysis with prod-code: 460,749 lines of Nim, multi-backend code generation, allocator test clone clusters, 1,824 invariant asserts, and a reported 94% AST slice reduction.

Analysis of Nim's package manager and build orchestrator with prod-code: 38,910 lines of Nim, a dependency scan that recognized no modules, SAT-based dependency resolution, and compiler AST declarative parsing.

Historical Nimbus consensus client analysis with prod-code: 179,711 lines of Nim, hard-fork specification clones, and 585 Result type matches.

Analysis of the D standard library with prod-code: 366,399 lines of D, a dependency scan that recognized no modules, range-based algorithms, Design by Introspection, and repository-wide matching-line counts for assertions and unittest occurrences.

Deep-dive analysis of phoenixframework/phoenix_live_view: 142,201 lines across 430 files, HEEx compile-time template splitting with runtime diffing, websocket state machines, 3,451 word-matching assertion lines, and 215 raise lines.

Deep-dive analysis of phoenixframework/phoenix: 100,976 lines across 508 files, generated route matching, channel multiplexing, and reproducible text-match counts.

Rails framework analysis with 3,478 Ruby files, Active Record queries, component packages, and test/error search results; subsystem line subtotals are withdrawn due to a discrepancy.

Analysis of the enterprise PHP component ecosystem created by Fabien Potencier with prod-code: 2,201,262 lines across 11,999 files, a 64-node zero-edge dependency extraction, ahead-of-time compiled DI passes, and 47,912 test assertion matching lines.

Analysis of the D asynchronous web framework with prod-code: 30,151 lines across 108 files, a two-node dependency extraction with no edges, compile-time REST interface generation, and repository-wide assertion matching lines.

Historical WordPress 6.8 source analysis with prod-code: 673,853 lines across 1,899 PHP files, WP_Hook dispatch, and a 693-node dependency extraction with no detected edges.

Historical ZLS analysis with prod-code: 56,190 lines of Zig, LSP protocol decoupling, completion test clone clusters, 86 errdefers, and an in-memory refactoring pre-flight check.

Moving a function to another module is cut and paste and then an unbounded amount of fixing every file that imported it. Building the tool that does both, and the two bugs that made it quietly skip the second.