# Search Log — EEWorld / 21ic / WeChat / Maker Spaces / Chinese engineer blogs

**Task**: Find Chinese engineers, makers, and small studios who can build a physical prototype of a wrist-worn pneumatic device (ESP32/STM32 + micro air pump + solenoid valve + pressure sensor + 3D-printed wrist housing + closed-loop pressure control).

**Search window**: 2026-06-30
**Tooling used**: `web_fetch` (direct HTTP), `exec`+curl (direct API), `web_search` (blocked — see constraint below).

---

## Environmental constraint: search engines are walled off

This run is happening from `claw.mni.ac` (egress `178.105.110.42`, hetzner falkenstein, AS24940). The IP range is on known datacenter lists that Chinese search engines gate with captcha walls. Concrete observed behavior:

| Engine | Result |
|---|---|
| DuckDuckGo | Returns `bot-detection challenge` HTML for every query. |
| Bing | Returns Turnstile captcha wall — `web_search` times out. |
| Google | Blocked behind `sorry/index` redirect. |
| Baidu | Direct curl returns the captcha page even with a normal UA. |
| Sogou WeChat | Captcha wall on first GET, also gated. |
| Brave Search | Returns search results but the results pages themselves are gated downstream. |

This means the canonical `site:eeworld.com.cn` / `site:21ic.com` / Sogou WeChat search strategy requested in the task **cannot be executed from this box**. Every search-engine fallback in the brief is unusable.

## What I did instead

I relied on three bypass routes:

1. **Direct platform visits** for sites where `web_fetch` returns content without a captcha: dfrobot.com, chaihuo.org, xinchejian.com, oshwhub.com (partial), hax.co.
2. **CSDN's internal search API** at `https://so.csdn.net/api/v3/search?q=…&t=blog` — this endpoint returns JSON without captcha and bypasses the search-engine wall. Discovered by accident; became the primary research route.
3. **Direct author profile fetches** for promising CSDN users (`blog.csdn.net/<username>`).

## Queries actually run (CSDN API)

The following queries were executed against `so.csdn.net/api/v3/search?q=…&t=blog`:

| # | Query | Top signal | Verdict |
|---|---|---|---|
| 1 | `微型气泵 单片机` | 特纳斯电子 airbag rehab rig | **Used as lead #1 (score 8)** |
| 2 | `气压 传感器 STM32 腕带` | 特纳斯电子 (same article) | Confirmed lead #1 |
| 3 | `微泵 Arduino 腕带` | generic Arduino pump tutorials | Student-project heavy, no usable lead |
| 4 | `穿戴 气压 单片机 ESP32 腕带` | 0 results | No overlap between wearables+气压+pump |
| 5 | `半导体刷机 腕带 气压` | 0 results | Nonsense query — discarded |
| 6 | `工作室 腕带气压 单片机 服务` | 上海依阳 pneumatic polishing | **Used as lead #5 (score 5)** |
| 7 | `微型 气泵 压力传感器 控制` | JavaSoul111 PID air-pump article | **Used as lead #2 (score 6)** |
| 8 | `腕充 气泵 压力传感器 控制` | refrigeration pressure-sensor consultant | **Used as lead #6 (score 4)** |
| 9 | `穿戴 压力传感器 腕充 ESP32` | 0 results | No overlap |
| 10 | `气压调节 腕带 腕充 气囊 压力` | 0 results | No overlap |
| 11 | `Seeed Studio 中国 穿戴 原型` | OpenGlass / XIAO ESP32S3 tutorials | **Used as leads #3, #4 (score 5 each)** |
| 12 | `腕充 气压 腕带 单片机 工作室` | 特纳斯电子 airbag rig | Confirmed lead #1 |
| 13 | `微型气泵 小泵 压力 腕带 制作` | refrigeration control articles | No new wearable builders |
| 14 | `微型气泵气泵 压力传感器 分体` | 0 results | Overly narrow |

**Key signal**: searches 4, 9, 10, 14 returned zero hits — there is essentially no public Chinese content combining "wearable + 气压 + pump + MCU". The intersection is so rare that finding any working build is meaningful.

## Channels that did NOT yield leads despite working fetches

- **EEWorld forum (`eeworld.com.cn`)** — Search-route URLs return 404; forum threads require session cookies. Could not search.
- **21ic forum (`21ic.com`)** — Same: search routes 404; forum threads behind session.
- **Sogou WeChat (`weixin.sogou.com`)** — Captcha wall on every request.
- **WeChat (`mp.weixin.qq.com`)** — No public search API.
- **Oshwhub (`oshwhub.com`)** — API endpoints return a CloudWAF "your request looks like an attack" 418 page on every curl. The site is reachable only through a real browser.
- **Zhihu (`zhihu.com`)** — `web_fetch` returns skeleton HTML; not enough content to evaluate.
- **立创开源硬件平台 (oshwhub.com)** — Same 418 wall.
- **Bilibili (`bilibili.com`)** — Out of scope per the task brief (not engineer blogs), but worth flagging as a future channel: there are likely Chinese wearable-pump prototype videos that I couldn't enumerate.
- **Hackaday.io** — Out of scope geographically.

## Platform leads (verified reachable, low signal individually)

These are communities/channels rather than direct leads — included in the CSV at score 3-5 to make the sourcing path explicit.

- **DFRobot (Shanghai/Beijing/Shenzhen)** — Maker platform with deep wearable tutorials. Contact via `info@dfrobot.com` or the forum. Score 5 as a sourcing channel.
- **柴火创客空间 (Chaihuo)** — Shenzhen makerspace, Seeed-affiliated. No engineer directory. Score 3.
- **新车间 (XinCheJian)** — Shanghai hackerspace. WeChat-driven. Score 3.
- **HAX** — Hardware VC accelerator. Not a prototyping service. Score 3 with caveat.

## Why so few leads at score 7+

The scoring rubric requires score 8-10 leads to show **public evidence of a pump + valve + sensor + MCU wearable build**. Of the 14 CSDN searches + 6 direct site visits, exactly **one author (tenas / 特纳斯电子)** publishes a build that is functionally equivalent to our device (airbag + pump + pressure sensor + MCU + closed-loop control). The next tier down — strong adjacent portfolios — has only a handful of entries. Most CSDN results for this query space are:

- 课程设计 / 毕设选题 (graduation-project kits) — not paid-engagement studios
- 制冷 / 空调 / 工业 pneumatic articles — relevant expertise but no wearable history
- 学生 blog posts (student projects) — portfolio-thin, not contact-ready

This means a score-8 lead is genuinely rare in the search space, not because I missed good leads but because the overlap "wearable + pump + pressure-sensor closed-loop" is genuinely a niche even in Chinese embedded blogs.

## Quality decisions

- **Conservative scoring**: Every lead carries a 1-line `note` that names the actual evidence behind the score, not boilerplate. Lead #2 (JavaSoul111) is scored 6, not 8, because the single PID pump-article doesn't prove wrist-wearable work even though it shows the right control loop.
- **Avoided hallucination**: Where searches returned zero usable results (queries 4, 5, 9, 10, 14), I did not invent leads to fill the spreadsheet. Empty results are an honest signal — they're documented in the query table above so the next person knows not to re-run them.
- **Out-of-scope profiles (deliberately excluded)**: 软件-only (CSDN authors whose work is MFC / SQL Server / MATLAB image-processing), big factories (大厂), component distributors. JavaSoul111 is included despite breadth because the relevant article is on-target.

## What would unblock more results

For the next iteration, the productive additions are:

1. **Run from a residential-IP browser** (not 178.105.110.42) to clear the search-engine captcha walls — opens up `site:eeworld.com.cn`, Sogou WeChat, Baidu, Zhihu proper.
2. **Bilibili search** — `bilibili.com/search?keyword=微型气泵+腕带` is likely to surface prototype-build videos with creator contact info.
3. **Oshwhub search** — only accessible via a real browser. The site has tagged project discovery for `智能硬件` and `医疗电子` tags.
4. **Hackaday Prize / Shenzhen Maker Faire speaker lists** — names of Chinese engineers who present at international events are likely to have English-language contact points.
5. **WeChat public account search** via Sogou — gated but contains high-signal Chinese-engineer project posts.

## Final deliverable

- **CSV**: `/home/claw/.openclaw/workspace/projects/pneumatic-wrist-prototype/leads-ee-misc.csv`
- **Leads**: 10 total (2 strong, 4 medium, 4 platform/channel entries)
- **Highest-confidence contact**: `tenas` (CSDN `blog.csdn.net/tenas`) — commercial prototyping studio in Shenzhen with public hardware-build evidence that matches our architecture.