When to open: You want to know what OpenAI / Anthropic / Google / Apple / Perplexity have done recently / their strategic inflection points / business model changes. This wiki organizes 71+ pieces on leading company moves from Shensi Circle by company.
One-sentence core:Anthropic's valuation doubled in three months surpassing OpenAI + moat = permissions not intelligence + Perplexity 450 million ARR (three and a half years / 5 major growth engines) + Sam Altman = Oppenheimer? + a16z's four major platform company playbooks (Google/Microsoft/AWS/Palantir) accelerate in the AI era.
0. Frequency of leading companies on Shensi Circle
| Company | Number of appearances | Key narrative |
|---|---|---|
| OpenAI | 19 | Sam Altman personally + GPT series + OpenAI fund + Sora + Operator |
| Anthropic | 10 | Self-growth black box + moat + finance + doubled valuation + Claude Code ambition |
| Perplexity | 6 | 450 million ARR + 5 major growth engines + business model |
| Multiple times (BERT/Gemini/AI Overviews) | Privacy Sandbox / control layer rewrite | |
| Apple | Multiple times | iMessage AI / iOS integration |
| Microsoft | Multiple times | Entra ID / Security Copilot |
| Meta | Few | Llama / VR |
| Karpathy | 2 | Software 3.0 / nanochat |
1. OpenAI (19 pieces)
1.1 Sam Altman's personal narrative
"10,000-word interpretation of OpenAI's founder: Is Sam Altman the Oppenheimer of our time?"
1.2 Sam's latest disclosures
"Sam Altman's latest disclosures: OpenAI's future plans"
1.3 OpenAI valuation timeline
- $500 billion → $830 billion (doubled in several months)
- "Kingmaking strategy" valuation surges in 27-60 days
1.4 OpenAI fund layout (vertical investments)
| Investment | Amount |
|---|---|
| AI+education Speak | $16 million |
| AI+automotive Ghost Autonomy | $5 million |
| AI applications (super cool Chinese founder) | $3.4 million |
1.5 OpenAI Operator / Sora / GPT series
- Sora Video (for detailed analysis see Shensi Circle / AI Industry Trends + New Categories + Forecasts)
- GPT 5 / 4o and other model iterations
1.6 Relationship with Anthropic
"Valuation doubled in three months surpassing OpenAI, yet Anthropic's old shareholders are waiting for IPO to cash out" "Two years of fighting over "who is the smartest" —— OpenAI and Anthropic's mutual rivalry
2. Anthropic (10 pieces)
See Shensi Circle / AI Agent Company Case Library §2:
2.1 Core proposition: moat = permissions not intelligence
(Jaya Gupta 2026-05)
"The scarcest thing in enterprise AI is shifting from "intelligence" to "permissions". "
2.2 Key action list
| Title | Highlights |
|---|---|
| 40+ people leverage 130 billion in revenue | "Self-growth black box" |
| 69 Anthropic employees use AI to do business | Internal dogfooding |
| Valuation doubled in three months surpassing OpenAI | Old shareholders waiting for IPO to cash out |
| Anthropic enters finance | Verticalization |
| 2025 overseas talent report 80% retention | Talent war |
| Marketing is dead, one person at Anthropic outperforms an entire marketing department | AI self-marketing (key case) |
| Why your Agent can't go to production | Anthropic's answer |
| Anthropic's "ambition" behind Claude Code | Skill / OS strategy |
| AI+recruitment: Inflection and Anthropic are already using it | Dogfooding case |
2.3 Key differences from OpenAI
- Anthropic: Builder-friendly (API + Skill + model quality)
- OpenAI: Consumer-friendly (ChatGPT + traffic)
- Result: Anthropic's valuation doubled in 3 months surpassing OpenAI
3. Perplexity (6 pieces)
3.1 Flagship: 450 million ARR (three and a half years)
"Doused with cold water by early investors on the spot, three and a half years later Perplexity reached $450 million ARR —— full breakdown of the five major growth engines"
(Shensi Circle in-depth article, see original document for details)
3.2 Perplexity CEO's business views
"Latest exclusive interview with Perplexity CEO: Most problems can be solved with internet search alone, a few require extra thinking; users don't care about your model or business approach"
3.3 Anti-framework technical inside story
"Revealing the technical inside story of Cursor, Perplexity, Lovable: Why they all chose the 'anti-framework' route"
3.4 Valuation positioning
"AI+search | Imagination about search, and Perplexity, the generative search engine with the highest current valuation"
3.5 New challenger
"AI+search | Former Baidu executive enters AI search, challenging Google and Perplexity, securing $60 million in seed funding"
4. Google
4.1 Privacy Sandbox case (cited from a16z Jaya Gupta)
"The company that built the tracking economy is now writing the rules after replacing the tracking economy."
Implication: Google is rewriting the control layer —— third-party cookies are blocked by regulators, and Google's own Privacy Sandbox API takes over.
4.2 Gemini / AI Overviews
See SEO × LLM Era (cross-author topic synthesis) + GSC Blog: AI Search Evolution and Google's Stance on AI Content (2015-2026)
4.3 NotebookLM
See RAG / Knowledge Base Engineering (cross-author topic synthesis) §End State (Obsidian + NotebookLM)
5. Apple
5.1 AI iMessage
"AI personal assistant moves into iMessage: two-step setup, $49/month, understands your schedule better than you do"
5.2 Apple Intelligence
See Wang Kai's AI Industry Judgments and Big Company Competitive Landscape Observations §Apple Evolution
5.3 Commercial war with OpenAI
See newtype AI Business, Industry and Investment Observations §OpenAI/Apple Commercial War
6. Microsoft
6.1 Entra ID + Security Copilot
(cited from a16z Jaya Gupta)
"Microsoft has layered another set on top of the same system of identity layer + Security Copilot: conditional access, PIM, identity threat detection, AI-driven security workflows."
"Owning the identity layer means owning the most valuable telemetry in the enterprise; owning telemetry means third-party security products structurally cannot beat your own Copilot. "
6.2 Relationship with OpenAI
- Microsoft holds a large stake in OpenAI
- Azure OpenAI is the enterprise-grade entry point
7. AWS
7.1 Selling infrastructure + guardianship together (a16z Jaya Gupta)
"AWS moves enterprise computing to Amazon's cloud, then starts selling the compliance, security, and governance tools enterprises need on the cloud.The more you move over, the less cost-effective it is to connect third-party security, so you might as well buy them together. "
7.2 AWS Agent series
See AWS Agent Series Lectures 2025 (4-piece compilation) —— 4-piece compilation (Wang Qixiang / Wang Xiaoyan / Chang Gaowei / idoubi)
8. Palantir
8.1 Workflow bundling (a16z Jaya Gupta)
"Palantir does not pursue volume, but goes deep into a small number of the most sensitive decision-making processes —— governments, large institutions. Once it becomes 'the place where decisions happen', it simultaneously becomes 'the place where decisions must be made securely'.These two things stick together into one, and can never be separated again. "
8.2 Rhythm comparison (compiled by a16z)
| Platform | Loop closure speed |
|---|---|
| AWS | Several years (infrastructure level) |
| Microsoft | Several months (identity level) |
| Palantir | Almost closes upon integration (workflow level) |
9. Karpathy (2 pieces / main index)
9.1 Karpathy's personal views
Shensi Circle cites relatively few of Karpathy's public speeches / X posts; for the in-depth wiki see Karpathy System (cross-author topic synthesis).
9.2 $100 million Series A (Fei-Fei Li + Karpathy)
"$100 million Series A! Fei-Fei Li and Karpathy bet big: This company will use AI to replicate all of humanity"
10. Meta / xAI / Others
10.1 Meta Llama series
- Open-source models → underlying choice for entrepreneurs
- See newtype Model Evaluation, Fine-tuning and Hardware Evolution
10.2 xAI
- Valuation $42.1 billion
- See AI Agent Market Investment Trends 2025 (data report)
10.3 Chinese leaders
- DeepSeek / Kimi / Zhipu GLM
- "When $50 billion Cursor runs ads for Kimi"(2026-03)
11. a16z four major platform company playbooks (key synthesis)
Shensi Circle / AI Agent Company Case Library §2.3 + extension of this section:
Old playbook (past 20 years):
- Google: Rewrite the control layer
- Microsoft: Lock in the identity layer
- AWS: Sell infrastructure + guardianship together
- Palantir: Bundle with scenarios until irreplaceable
New playbook (accelerated in the AI era):
- Capability as threat(Anthropic model release original quote: "Improvements that patch vulnerabilities also make it stronger at exploiting vulnerabilities")
- The more it is used, the better it understands you —— AI learning to "organize context" becomes a new moat
"The company that best understands what this model can do in your company is also the company that best understands how this model can in turn harm you. They are the same company."
12. How to use —— leading company action paths by company
| You want to understand | Main use | Starting point |
|---|---|---|
| OpenAI strategy | §1 Sam Altman + OpenAI fund + Operator | "Oppenheimer" narrative + vertical investments |
| Anthropic moat | §2 Jaya Gupta + self-growth black box | Permissions > intelligence + AI self-marketing |
| Perplexity growth | §3 450 million ARR + 5 major engines | Three and a half years / CEO business views / anti-framework |
| Google AI strategy | §4 Privacy Sandbox + Gemini | Control layer rewrite |
| New platform company playbook | §11 compiled by a16z | Old playbook + acceleration in the AI era |
| OpenAI vs Anthropic | §2.3 key differences | Builder vs Consumer |
General discipline:
- Doubling in three months(Anthropic valuation) is the norm for leading companies
- Moat = permissions > intelligence > context(2026 latest consensus)
- Platform company playbook accelerates in the AI era(Capability as threat + the more used the better understood)
- OpenAI Consumer / Anthropic Builder divergence
- China-US leading company standoff: Cursor running ads for Kimi = global cooperation regardless of stance
13. Citation list
- Shensi Circle full author collection entry —— full collection entry
- Main sources:
sources/learning/深思圈/~71+ articles covering "OpenAI" / "Anthropic" / "Google" / "Apple" / "Perplexity" / "Microsoft" and more - AI Industry Judgments and Corporate Competition (Cross-Author Thematic Synthesis) —— Comprehensive Industry Judgments
- newtype Observations on AI Business, Industry and Investment —— newtype Perspective (Battle of Three Camps / Commercial Warfare)
- Wang Kai's Observations on AI Industry Judgments and the Competitive Landscape of Large Companies —— Judgments from Wang Kai's 790 Posts
- Shensi Circle / AI Agent Company Case Library —— Details of Anthropic's Moat
- Karpathy System (Cross-Author Thematic Synthesis) —— Karpathy System
Sources and Related Materials
- https://mp.weixin.qq.com/s?__biz=MzAxNjE1NTg2Ng==&mid=2247494188&idx=1&sn=394115e9006df186ed1c2944a1aa01b8&chksm=9bfbbf99ac8c368fb1821f0d52491195fcaaf55d2c6557c7c53c6621674a9840ece831b72246
- Full Entry Portal for Shensi Circle Authors
- Shensi Circle / AI Primary Market VC Perspective (a16z + Sequoia + YC)
- Shensi Circle / AI Agent Company Case Library
- Shensi Circle / AI Industry Trends + New Categories + Forecasts
- AI Industry Judgments and Corporate Competition (Cross-Author Thematic Synthesis)
- newtype Observations on AI Business, Industry and Investment
- Wang Kai's Observations on AI Industry Judgments and the Competitive Landscape of Large Companies