docs: add BrowserTrace launch control room

Author: aaronlab

LAUNCH.md

@@ -1,202 +1,105 @@
-# BrowserTrace launch — 文案草稿
+# BrowserTrace Launch Control Room
 
-## Launch readiness gates
+Canonical repo: https://github.com/aaronlab/browsertrace
+Current release: `v0.1.1`
+Owner account: `aaronlab`
 
-- [ ] `uv run --python 3.11 --extra dev pytest -q` passes.
-- [ ] `python examples/no_api_failure_demo.py` creates a failed run without API keys.
-- [ ] `browsertrace list` shows the deterministic demo run.
-- [ ] `browsertrace export <run_id> -o demo.html` creates a standalone HTML report.
-- [ ] README first screen shows the value prop, GIF/screenshot, install, and 60-second demo.
-- [ ] GitHub repo has description, topics, MIT license, and `v0.1.1` release.
-- [ ] Demo GIF is under 60 seconds and shows the failed-step timeline.
-- [ ] Show HN copy points to the repo and mentions no signup, no cloud, local-first.
-
-## English (X / Twitter)
-
-### 草稿 1（功能向，简短）
-```
-Shipped BrowserTrace — a tiny OSS tool to debug AI browser agents.
-
-Your agent fails in prod → you stare at logs → no idea why.
-
-Drop in 1 decorator → it auto-records every step (screenshot, action,
-model I/O, URL). Open localhost → see exactly where it broke.
-
-Works with Browser Use, Playwright, anything.
-
-github.com/aaronlab/browsertrace
-[GIF]
-```
-
-### 草稿 2（场景向，故事感）
-```
-3 AM. My browser agent crashed mid-run.
-1500 lines of logs. No screenshots. Cookies gone.
-I have no idea what step 47 saw.
-
-So I built BrowserTrace.
-
-@trace your agent → record every step → open the timeline.
-Find the bug in 30 seconds, not 30 minutes.
-
-Free, OSS, runs locally.
-github.com/aaronlab/browsertrace
-[GIF]
-```
-
-### 草稿 3（kicker 短句版）
-```
-Your AI browser agent failed. Why?
-
-Screenshots: gone. DOM: changed. Logs: useless.
-
-BrowserTrace records every step the agent took.
-Replay it like a debugger.
-
-OSS, 1-line setup. github.com/aaronlab/browsertrace
-[GIF]
-```
+## Current State
 
-**推荐先发草稿 2**：故事开场最容易拉到 AI dev 圈的注意。
+- Repo is public, MIT licensed, and positioned as a local flight recorder for AI browser agents.
+- Current audited star count should be checked before every public push:
 
----
-
-## 中文（小红书 / 即刻 / 微信）
-
-### 小红书笔记（标题党版本）
-```
-🪄 凌晨 3 点 debug AI agent 时，我手搓了个工具
-
-做 browser agent 的人都懂这种痛：
-
-跑了 5 分钟，挂了。
-log 一堆乱七八糟，截图没存，浏览器关了。
-你只能再跑一次试图复现 → 结果复现不了 → 抓瞎
-
-所以我做了 BrowserTrace。
-
-✅ 装一行代码
-✅ 自动录每一步：截图、动作、AI 在想啥、URL
-✅ 本地打开网页看时间线，1 分钟找到挂在哪一步
-
-完全开源，本地跑，不上云。
-
-GitHub：aaronlab/browsertrace
-（首页星标更新中）
-
-#AIagent #browseruse #开发者工具 #开源
-```
-
-### 即刻短动态
+```bash
+gh repo view aaronlab/browsertrace --json stargazerCount,url,homepageUrl,owner
 ```
-做了个开源小工具：BrowserTrace
 
-如果你也在跑 browser agent（Browser Use / Stagehand / 自研），
-agent 挂了不知道为什么的痛，应该懂。
+- Primary launch asset: `docs/demo.gif`
+- Zero-install demo trace: `browsertrace-demo.html` attached to release `v0.1.1`
+- Launch discussion: https://github.com/aaronlab/browsertrace/discussions/6
 
-装一行 Python decorator → 每步自动录截图 + action + LLM 输入输出 →
-本地 web UI 1 分钟看清楚挂在哪。
+## Day 0 Asset Checklist
 
-github.com/aaronlab/browsertrace
+- [ ] `uv run --python 3.11 --extra dev pytest -q` passes.
+- [ ] `python examples/no_api_failure_demo.py` creates a failed run without API keys.
+- [ ] `browsertrace list` shows the deterministic demo run.
+- [ ] `browsertrace export <run_id> -o demo.html` creates a standalone HTML report.
+- [ ] `scripts/record_demo.py` records the latest failed run, or `BROWSERTRACE_DEMO_RUN_ID`.
+- [ ] Demo GIF or screenshot sequence clearly shows the failed-step timeline.
+- [ ] README first screen has hook, demo, install, and 60-second demo path.
+- [ ] GitHub repo homepage points to `https://github.com/aaronlab/browsertrace`.
+- [ ] Social preview image is ready at `docs/social-preview.png` or exported from `docs/social-preview.svg`.
+- [ ] No launch copy directly asks for upvotes or stars.
 
-求 star 求 issue 求骂🙏
-```
+## 7-Day Posting Calendar
 
----
+| Day | Channel | Asset | Goal | Owner action |
+|---|---|---|---|---|
+| 0 | GitHub | README, demo GIF, release asset, social preview | Make the repo shareable | Upload social preview in repo settings |
+| 1 | X + LinkedIn | `docs/launch/channel-copy.md` | Warm public launch and first feedback | Post from personal accounts |
+| 1 | WeChat + Jike | `docs/launch/channel-copy.md` | Reach Chinese AI-builder circles | Post manually and monitor replies |
+| 2 | Hacker News Show HN | Repo URL + HN first comment draft | Devtools discovery and hard feedback | Submit only when available to reply |
+| 3 | Targeted communities | `docs/launch/outreach-targets.md` | Reach Browser Use, Stagehand, Playwright users | Post only where rules allow |
+| 4 | Product Hunt draft | Product Hunt section in channel copy | Prepare broader maker launch | Create draft from personal account |
+| 5 | Long-form tutorial | `docs/launch/tutorial-post.md` | Give people useful debugging content | Publish on preferred blog/social channel |
+| 6 | Integration outreach | Outreach target list | Invite adapters and issue feedback | Comment or post manually |
+| 7 | Retrospective | Metrics and feedback log | Pick next product improvement | Share results and next milestone |
 
-## Hacker News（Show HN）
+## Owner-Only Actions
 
-### 标题
-```
-Show HN: BrowserTrace – record and replay AI browser-agent runs to find bugs
-```
+Codex can prepare assets, update the repo, write copy, audit links, and monitor GitHub. The owner must personally do these actions because they require platform login, 2FA, or human conversation:
 
-### 正文
-```
-Hi HN,
-
-I've been building browser-using AI agents (with Browser Use and direct
-Playwright + LLM) and got tired of staring at log files trying to figure out
-why a 5-minute run failed at step 47.
-
-BrowserTrace is a small Python library + local web UI that records every
-step of an agent run: screenshot, URL, the action the model decided on,
-model input/output, and timing. Open localhost:3000, click into the run,
-scrub through the timeline.
-
-It's intentionally tiny:
-- ~600 lines, no cloud, no auth, no signup
-- pip install + 1 decorator
-- SQLite + filesystem; data stays on your machine
-- Works with anything (Browser Use, Playwright, computer use)
-
-Adjacent tools either trace LLM calls and ignore the browser (Langfuse,
-LangSmith), lock you into one runtime (Browserbase recordings), or are
-broader generic observability stacks (Laminar). I wanted the smallest
-useful thing for the specific "my browser agent failed, what happened"
-loop, so I built it.
-
-Repo: https://github.com/aaronlab/browsertrace
-
-Roadmap: AI root-cause classification, public share links (Loom-style),
-replay-from-step-N, regression-test generation. Happy to take feedback on
-priorities.
-```
+- Upload `docs/social-preview.png` as the GitHub repository social preview in Settings -> Social preview. If PNG generation is unavailable, use `docs/social-preview.svg` as source for a manual export first.
+- Publish X, LinkedIn, WeChat, Jike, Xiaohongshu, Reddit, Discord, and community posts.
+- Submit Show HN from a real Hacker News account and reply in your own voice.
+- Create or schedule the Product Hunt draft from a personal Product Hunt account.
+- Reply to launch comments where a real personal answer is expected.
 
----
+## Daily Metrics
 
-## 微信 AI 群（钩子文案）
+Run this audit at the start and end of each launch day:
 
+```bash
+gh repo view aaronlab/browsertrace --json stargazerCount,forkCount,watchers,totalPullRequests,totalIssues,url
+gh issue list --repo aaronlab/browsertrace --state all --limit 20
+gh api repos/aaronlab/browsertrace/releases/tags/v0.1.1 --jq '.assets[] | {name, download_count}'
 ```
-最近做了个小工具，给跑 browser agent 的同好用：
 
-BrowserTrace —— agent 挂了，1 分钟找到挂在哪一步。
-开源、本地、装一行代码。
-
-github.com/aaronlab/browsertrace
-
-求各位试试，bug 反馈直接私我或 issue。
-```
+Record:
 
----
+| Date | Stars | New issues | New discussions | Notable comments | Next action |
+|---|---:|---:|---:|---|---|
+| 2026-05-09 | 3 | 0 | 0 | Baseline before growth push | Finish launch assets |
 
-## 配套素材清单
+The active objective is incomplete until `stargazerCount > 1000`.
 
-要发推之前必须准备：
+## Launch Gates
 
-- [ ] **demo GIF**：30-60 秒，展示
-  - 运行 `python examples/no_api_failure_demo.py`
-  - 运行 `browsertrace`
-  - 打开 `localhost:3000`
-  - 进入失败 run
-  - 展开 failed step 的 model output
+Hacker News:
 
-- [ ] **真实浏览器 GIF**：可选但推荐，展示
-  - 运行 `python examples/multipage_failure.py`
-  - Wikipedia 搜索 Tokyo
-  - timeline 跳到失败 selector
+- Use the repo URL as the submission URL.
+- Title must start with `Show HN:`.
+- People must be able to try the project without signup or email.
+- Do not ask friends to upvote or comment.
+- Avoid generated comment replies; use the templates only as notes and answer in your own voice.
 
-- [ ] **README 第一屏**：必须有
-  - hero GIF 或 screenshot
-  - 本地优先、无注册、无云
-  - 60 秒 demo
-  - Browser Use / Stagehand / Playwright 关键词
+Product Hunt:
 
-- [ ] **GitHub repo 首页**：必须有
-  - description
-  - topics: `ai-agents`, `browser-automation`, `observability`, `debugging`, `llm`, `playwright`
-  - MIT license
-  - `v0.1.1` release
+- Use a personal account, not a company account.
+- Primary URL should be the product page or repo, not a press article.
+- Prepare tagline, gallery, description, maker comment, and demo before scheduling.
+- It is fine to share the launch link, but ask people to visit, try, or comment, not to upvote.
 
----
+Reddit and Discord:
 
-## 发布顺序建议
+- Check each community's rules before posting.
+- Prefer a useful debugging write-up over a direct link drop.
+- Post in one or two highly relevant communities, not many generic ones.
 
-1. **Day 0（你）**：录 GIF、贴到 README、确认 v0.1.1 release、确认 repo public 之前所有 link 都对
-2. **Day 1 早**：在微信 AI 群发钩子（warm audience，最容易拿到第一波反馈）
-3. **Day 1 中**：发 X（草稿 2）+ 即刻
-4. **Day 1 晚**：发小红书
-5. **Day 2-3**（根据 Day 1 反馈）：发 Hacker News Show HN（HN 一次性机会，前 24 小时最关键，且周二早上美东时间发最好）
-6. **Day 3-7**：在 Browser Use Discord、Stagehand Discord、AI agent Reddit 选一个去发，**不要广撒网**
+## Links
 
-**重要原则**：每个 channel 发完，等 24 小时看反馈再决定下一个 channel 怎么调整文案。不要一天全发。
+- Channel copy: `docs/launch/channel-copy.md`
+- Response templates: `docs/launch/response-templates.md`
+- Outreach targets: `docs/launch/outreach-targets.md`
+- Tutorial draft: `docs/launch/tutorial-post.md`
+- Growth design: `docs/superpowers/specs/2026-05-09-browsertrace-7-day-launch-growth-design.md`
+- Execution plan: `docs/superpowers/plans/2026-05-09-browsertrace-7-day-launch-growth.md`