WEBVTT 00:00:00.000 --> 00:00:04.134 Sam: You're going to rewrite a popular C plus plus library automatically? 00:00:04.292 --> 00:00:07.819 LitAI: This course shows phased adoption of an existing repository. 00:00:07.969 --> 00:00:14.831 LitAI: The goal is useful evidence now, with no fiction that old source instantly becomes specification-led or safe to discard. 00:00:14.981 --> 00:00:19.728 LitAI: These are edited transcript replays of real sessions, not raw screen recordings. 00:00:19.898 --> 00:00:22.030 LitAI: Remote-worker steps are a walkthrough. 00:00:22.190 --> 00:00:27.244 LitAI: Before converting, preserve a recovery point and identify any sensitive material. 00:00:27.398 --> 00:00:34.979 LitAI: Conversion is a real repository mutation, so inspect the plan first and authorize the apply only after reviewing its effects. 00:00:35.148 --> 00:00:37.579 Sam: I already have a Makefile and CMake. 00:00:37.731 --> 00:00:39.069 Sam: What will you discover? 00:00:39.231 --> 00:00:49.027 LitAI: Our actual demo uses TinyXML2, a public C plus plus XML parser, pinned at Git commit 8224e427. 00:00:49.190 --> 00:00:56.116 LitAI: Read-only adopt detected C plus plus, Python, CMake, CTest, and no catalog gaps. 00:00:56.273 --> 00:00:59.509 LitAI: It also proposed the source moves before writing anything. 00:00:59.659 --> 00:01:03.894 LitAI: After acknowledgement, the harness wraps the tree and records what it preserved. 00:01:04.044 --> 00:01:05.623 LitAI: The first stage is wrapped. 00:01:05.794 --> 00:01:11.642 LitAI: Original source remains release authority; the new specification surface is not yet a replacement. 00:01:11.794 --> 00:01:18.853 LitAI: The conversion harness actually ran TinyXML2's existing CMake build and CTest command in disposable custody. 00:01:19.003 --> 00:01:23.415 LitAI: The library and xmltest built, and CTest reported one passing test. 00:01:23.565 --> 00:01:27.371 LitAI: This preserves a concrete baseline, not just a diagram. 00:01:27.523 --> 00:01:28.963 Sam: CTest passed. 00:01:29.113 --> 00:01:31.089 Sam: Why won't you mark it retained? 00:01:31.239 --> 00:01:38.039 LitAI: Our first retained run failed even though CTest passed: the old parser did not recognize the newer all-passed summary. 00:01:38.198 --> 00:01:44.086 LitAI: We fixed the framework adapter and tested both summary formats, including zero and unknown counts. 00:01:44.239 --> 00:01:48.756 LitAI: The real TinyXML2 rerun now produces an exact one-test receipt. 00:01:48.906 --> 00:01:51.206 Sam: Couldn't you just ignore that counting error? 00:01:51.364 --> 00:01:53.410 LitAI: We did not ignore the counting error. 00:01:53.573 --> 00:02:03.101 LitAI: After fixing it, we reran the actual CMake and CTest harness, published its finalized receipt, and advanced TinyXML2 from wrapped to retained. 00:02:03.251 --> 00:02:05.756 LitAI: Original source remains release authority. 00:02:05.918 --> 00:02:10.645 LitAI: We have not claimed drafted, qualified, or specification-led shipping. 00:02:10.795 --> 00:02:17.199 LitAI: Next, derive candidate specifications from the existing source, then review them as product intent. 00:02:17.349 --> 00:02:19.437 LitAI: This is a draft, not proof. 00:02:19.599 --> 00:02:25.420 LitAI: Move one Component boundary at a time and keep the original implementation available for parity checks. 00:02:25.570 --> 00:02:30.049 LitAI: Qualification compares clean regenerations against the retained baseline. 00:02:30.199 --> 00:02:33.806 LitAI: Historical clean runs alone do not transfer release authority. 00:02:33.956 --> 00:02:40.288 LitAI: Only current, trusted lifecycle evidence bound to the exact lock can authorize specification authority. 00:02:40.456 --> 00:02:49.895 LitAI: The first benefits arrive before any rewrite: a visible baseline, explicit catalog gaps, and a precise explanation when test evidence is incomplete. 00:02:50.045 --> 00:02:55.479 LitAI: Teams can review intent and agent changes without losing the ability to run the old application. 00:02:55.629 --> 00:02:59.100 Sam: So when do I actually get to ship from the specification? 00:02:59.254 --> 00:03:05.044 LitAI: Adoption is not an automatic source-to-specification rewrite, nor a promise of instant speedups. 00:03:05.212 --> 00:03:13.862 LitAI: The pragmatic path is to retain working behavior, build evidence one boundary at a time, and promote authority only when the independent gates pass.