Deep Dive into LLMs like ChatGPT
说话人分离结果（匿名标签）。同目录的
transcript.md未作任何改动。 标签 → 真人姓名的映射属于编辑判断，请记在 wiki 视频页开头，见 CLAUDE.md。 分离在抢话/串场处会错配：标签是证据不是判决，与内容矛盾时以内容为准。
说话占比¶
| 说话人 | 时长 | 占比 |
|---|---|---|
| SPEAKER_00 | 03:29:52 | 100.0% |
| SPEAKER_01 | 00:00:02 | 0.0% |
（访谈里占比低的通常是主持人，可作为认人的线索之一。）
说话轮次¶
| 起 | 止 | 说话人 |
|---|---|---|
| 00:00:00 | 00:03:45 | SPEAKER_00 |
| 00:03:48 | 00:13:20 | SPEAKER_00 |
| 00:13:23 | 00:13:37 | SPEAKER_00 |
| 00:13:42 | 00:41:09 | SPEAKER_00 |
| 00:41:13 | 00:45:32 | SPEAKER_00 |
| 00:45:34 | 00:48:27 | SPEAKER_00 |
| 00:48:31 | 00:48:42 | SPEAKER_00 |
| 00:48:46 | 00:49:19 | SPEAKER_00 |
| 00:49:24 | 00:55:32 | SPEAKER_00 |
| 00:55:34 | 00:58:35 | SPEAKER_00 |
| 00:58:38 | 00:59:30 | SPEAKER_00 |
| 00:59:33 | 01:09:42 | SPEAKER_00 |
| 01:09:32 | 01:09:33 | SPEAKER_01 |
| 01:09:42 | 01:18:45 | SPEAKER_00 |
| 01:18:50 | 01:25:09 | SPEAKER_00 |
| 01:25:12 | 01:25:20 | SPEAKER_00 |
| 01:25:23 | 01:26:23 | SPEAKER_00 |
| 01:26:25 | 01:29:44 | SPEAKER_00 |
| 01:29:46 | 01:29:49 | SPEAKER_00 |
| 01:29:52 | 01:30:06 | SPEAKER_00 |
| 01:30:09 | 01:30:22 | SPEAKER_00 |
| 01:30:25 | 01:30:26 | SPEAKER_00 |
| 01:30:28 | 01:30:32 | SPEAKER_00 |
| 01:30:35 | 01:56:48 | SPEAKER_00 |
| 01:56:51 | 02:04:40 | SPEAKER_00 |
| 02:04:42 | 02:05:53 | SPEAKER_00 |
| 02:05:57 | 02:26:20 | SPEAKER_00 |
| 02:26:02 | 02:26:02 | SPEAKER_01 |
| 02:26:21 | 02:46:21 | SPEAKER_00 |
| 02:46:24 | 02:54:13 | SPEAKER_00 |
| 02:54:11 | 02:54:12 | SPEAKER_01 |
| 02:54:13 | 02:58:02 | SPEAKER_00 |
| 02:58:04 | 02:59:08 | SPEAKER_00 |
| 02:59:10 | 03:09:35 | SPEAKER_00 |
| 03:09:32 | 03:09:33 | SPEAKER_01 |
| 03:09:35 | 03:13:12 | SPEAKER_00 |
| 03:13:11 | 03:13:11 | SPEAKER_01 |
| 03:13:13 | 03:13:48 | SPEAKER_00 |
| 03:13:43 | 03:13:43 | SPEAKER_01 |
| 03:13:48 | 03:15:22 | SPEAKER_00 |
| 03:15:24 | 03:25:59 | SPEAKER_00 |
| 03:26:01 | 03:31:21 | SPEAKER_00 |
| 03:31:20 | 03:31:20 | SPEAKER_01 |