feat(video): 《上线之后，AI 才开始上学》经验时代综述科普视频工程 + 公共视频制作管线沉淀 - #1102
Merged
ThreeFish-AI merged 5 commits intoAug 17, 2026
Conversation
- 新增 media/pipeline/：中心脚本三件套（build_narration/tts/qa_frames，--project 参数化，tts 增 --voice/--rate）+ Pipeline 总纲 README（九阶段 Stages、两种工程模式、复用边界、新集脚手架清单）+ skills/01-05 各阶段代理提示词规格； - 上集 scripts/*.py 改为薄包装转发 ../../pipeline/scripts/，保持原 CLI 契约；改造后 narration.json 字节级等价验证通过（shasum 一致）； - qa_frames CLI 规整：--scene 与句 id 位置参数互斥二选一。 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
- 基于清华×Frontis.AI 88 页综述《Self-Improving Agents in the Era of Experience》制作，与上集《AI 如何自己变强？》互补（改什么 vs 怎么攒经验），片尾互相引用； - research/paper-notes.md：9 个并行提取代理逐章精读产出的单一事实源（四段式：主旨/代表方法/风险防护/科普素材），IEEE 引用 + 140 号工程调研交叉引用； - script/：七幕策划案（金=经验流/青=Harness/紫=参数内化三色契约）+ 172 句逐字稿（12.6 分钟纯语音）+ 分镜表； - video/：Remotion 独立工程，七幕场景组件（P0Hook–P6Ending），manifest 驱动音画自动对轨，typecheck 零错误，草渲抽帧 QA（30+ 帧/7 幕目检）通过后终渲 1080p30（13.6 分钟，out/final.mp4 本地产物不入 git）； - 根 .gitignore 追加本集产物规则；knowledge-map 增补「科普视频制作 Pipeline」条目串联基建与两部作品。 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
- 3-E 自指套娃：百分比定位量纲错配致整幕空白，改 px 居中；
- 九处底部公式角标落入字幕条区域被 68% 黑底压暗，统一抬高至 bottom ≥ 150；
- 1-F 快慢弧线：pathLength=1 与像素级 dasharray 量纲互相抵消致渐进动画失效，改用 mask 分离绘制进度与虚线样式；
- 4-D 纵向时间线：删除多余 pathLength={1} 恢复虚线外观；
- 6-E 片尾渐黑窗口 [220,260] 超出 beat 实际帧数被硬切，改为从末句实时长推导 endFrame 收进 beat 内。
全部修复经 Remotion still 全分辨率抽帧像素级验证。
🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist)
Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
- 开屏英文引语换为 Silver & Sutton 原文逐字句（摘要句）， 并在 paper-notes.md 补录 Era of Experience 原文锚点； - p5-19 口播计数断档改为「更阴的还有一招」， 重生成 narration.json 并增量重合成该句 TTS； - storyboard.md 与场景实现对齐：2-D 并入 2-C（p2-07..12）、 后续镜号顺移，P5 插入 5-A2 引子行、5-B 改 p5-08..12， 同步代码 Sequence 名称； - 金句卡 ASCII 直引号改直角引号「」； - 删除模板遗留死代码（ChapterCard 组件、 FourDestinationsPreview.highlight、Ladder.lit）； - pipeline/README.md 脚手架清单 git cp 改为 cp -r。 🤖 Generated with [Claude Code](https://github.com/claude), [CodeX](https://openai.com), [Gemini](https://github.com/apps/gemini-code-assist) Co-Authored-By: Aurelius Huang<threefish.ai@gmail.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
概要
基于清华 × Frontis.AI 88 页综述《Self-Improving Agents in the Era of Experience: A Survey of Self- to Meta-Evolution》（2026-06）制作第二期科普视频，并将两期视频的制作流程沉淀为全仓可复用的公共 Pipeline。
一、公共 Pipeline（media/pipeline/）
沿用上集（PR #1101）验证过的全链路，抽出可复用基建：
--project参数化，任何media/*-video/工程可用）：build_narration.py：narration.md → narration.json + 时长估算tts.py：逐句 edge-tts 合成 + 时长 manifest（幂等；新增--voice/--rate）qa_frames.py：按句 id / 按幕抽帧视觉 QA（--scene与 ids 二选一）narration.json字节级等价验证通过（shasum 一致）二、新一集《上线之后，AI 才开始上学》（media/experience-era-agents-video/）
research/paper-notes.mdscript/planning.mdscript/narration.mdscript/storyboard.mdvideo/out/final.mp4内容主线（与上集互补：上集讲「自我进化改什么」，本集讲「部署之后经验怎么攒」，片尾互相引用）：
P0 经验时代宣言 → P1 部署态 AI 解剖（A_t=⟨M,H,U,E⟩、trace→经验提纯、三代演进）→ P2 经验四个去处（技能库/记忆/环境/参数巩固）→ P3 元进化三级阶梯（SkillOS 分岗 / Hyperagents 自指）→ P4 SI 六评测目标 + SIP-Bench → P5 移动攻击面（ClawHavoc ~1200 恶意技能 / 记忆投毒 >90% / 四味药）→ P6 开放问题收尾。
验证
narration.json薄包装改造前后 shasum 一致（字节级等价）tsc --noEmit零错误🤖 Generated with Claude Code
Co-Authored-By: Aurelius Huangthreefish.ai@gmail.com