Shensi Quan Author Complete Works Portal

One-sentence positioning

A Chinese-language self-media focused on AI entrepreneurship / company case studies / venture capital observations, has published simultaneously on three platforms since 2022, and reached near-daily update density in 2025-2026 (270+ articles per year). It covers the core moves of leading AI companies such as a16z / Sequoia / Anthropic / OpenAI / Cursor / Perplexity,1–2 weeks earlier than other domestic self-media to provide in-depth analysis.

When to come here to look things up

• "What do a16z / Sequoia / YC think about AI recently?" → grep a16z / 红杉 / YC in sources
• "How much did this AI company (Cursor / Anthropic / Perplexity / Notion / Figma) recently raise / how do they do growth?"
• "What are valuations like in the AI Agent track right now / which companies are working on it / which ones have died?"
• "Want to see the 0-1 details of a specific AI startup (fundraising, PMF, growth engine)"
• "What did Karpathy / Reid Hoffman / Sam Altman say recently that is worth noting down"
• Want "foreign in-depth articles → Chinese version" to save translation costs (the author does secondary processing on every piece + adds personal opinions)

Data panorama

Source distribution

Deduplication finding: about 25% of articles were published on two platforms simultaneously,and have been automatically normalized(the more complete version is retained; duplicate versions are in _archive_dup_2026-05-17/).

Time distribution

graph LR
  Y22["2022<br/>21 篇<br/>试水期"] --> Y23["2023<br/>89 篇<br/>加速"] --> Y24["2024<br/>116 篇<br/>稳定"] --> Y25["2025<br/>276 篇<br/>日更密度"] --> Y26["2026<br/>157 篇<br/>持续"]
  classDef hot fill:#fff3e0,stroke:#fb8c00
  class Y25,Y26 hot

Explosion point 2025: The AI Agent boom began, and the author clearly accelerated the pace. 5–6 articles per week became the norm.

Topic density (matched by title keywords)

The three keywords "US dollars + financing + a16z" account for 1/3 of the articles — this author's core perspective is VC + company actions, not technical depth nor PM/user perspective. He views AI through the lens of "which company is worth investing in / whether the valuation is reasonable / whether the founder is reliable".

Author profile (reverse-inferred from content)

Differences between the three accounts

The two WeChat accounts have almost equal article counts (236/235), but their topic focuses are slightly different:

• Shensi SenseAI: A more updated "product + technology" perspective, started in 2024-03, density increased from 2025 onward
• Shensi Quan (this account): Early account, started in 2022, retains some earlier "entrepreneurship stories" and "personal reflections"
• woshipm: 100% reposted from WeChat, is a public channel (searchable without following the WeChat account)

If you only want to read "the best pieces from the last year",prioritize SenseAI; if you want to see how the author thought in the early days, go to this account; if you want content that is search-engine indexable, go woshipm.

Index (by platform + year-month)

Directly grep ~/kb/sources/learning/深思圈/:

sources/learning/深思圈/
├── woshipm/
│   ├── 2022-06/ → 2026-05/    (188 .md / 1656 imgs)
│   └── ...
├── weixin/
│   ├── 2024-03/ → 2026-05/    (471 .md / 3831 imgs)
│   └── ...
├── _archive_dup_2026-05-17/   (163 dup, 30 天后可清)
└── _meta/
    ├── FINAL_REPORT.md         (爬取报告)
    ├── woshipm_manifest.jsonl
    ├── weixin_manifest.jsonl
    └── ...

Every source is structured.md:

• frontmatter contains title / url / publish_time / cover / description / tags
• Main text follows chapter structure (h2/h3)
• Images localized (relative path references)

Cross-wiki links (known intersections)

This batch has significant overlap with the existing wiki on three major themes, and it is recommended to merge them when extracting topic wikis in the future:

Note: This is an entry wiki,no in-depth topic merging has been done. When one day you ask "what are AI Agent valuations like now" / "which AI track is a16z betting heavily on recently", then trigger Phase 1 to extract a specific topic wiki (at that time it will be merged across authors, following Path B of SCHEMA).

Scraping technical notes (for traceability only)

• Systematic scraping rules: wiki-weixin-mp-scraping (documented in memory) + wiki-yitang-scraping
• Chrome MCP sandbox: see wiki-chrome-mcp-pitfalls
• Full report:sources/learning/深思圈/_meta/FINAL_REPORT.md

Known limitations

• 60 articles were deleted by the author: cannot be scraped, marked in manifest as fail / reason: "no content". If one day you search for a specific topic and cannot find it, it may have been deleted
• SenseAI account articles before mid-2025: only fully retrievable via the fakeid interface, not visible on wewe-rss (already worked around)
• Account B "Shensi Quan" actually has 405 entries: including multiple pushes on the same day (headline + sub-headline), not 405 independent articles
• "QR codes / promotional cards" at the end of WeChat Official Account articles: may cause the WeChat version to be misjudged as "more complete" during deduplication; this has been downweighted in the scoring logic

Subsequent paths