Articles Collection

Blog posts, articles, long reads, and educational content.

Recently Added

• Harper Carroll AI Course — Session 1: Foundations of Generative AI (2026-02-21) — Mental model of how modern generative AI works, from tokenization + embeddings to RAG and agent patterns.
• Harper Carroll AI Course — Session 2: The Method (2026-02-21) — Practical prompting: TCFCE framework, reasoning models, and the sycophancy problem.
• A Design Thinker's Guide to AI and Creativity (2026-02-18) — Jeremy Utley (Stanford d.school) on treating AI as a teammate, not a tool.
• How Will OpenAI Compete (2026-02-19) — Ben Evans: OpenAI's 4 strategic problems — no tech moat, shallow engagement, no PMF-owning products, no roadmap control.

Format

## Article Title
**Author:** [Name] | **Source:** [Publication](URL) | **Date Added:** YYYY-MM-DD
**Tags:** tag1, tag2, tag3

**Summary:** Brief overview of the main points.

**Key Takeaways:**
- Important point 1
- Important point 2

**Notes:** Personal thoughts or how it applies.

---

Amazon Earnings, CapEx Concerns, Commodity AI

Author: Ben Thompson | Source: Stratechery | Date Added: 2026-02-10 Tags: tech, finance, amazon, aws, ai, capex, cloud

Summary: Ben Thompson analiza los resultados trimestrales de Amazon y su anuncio de $200B en CapEx para 2026 — un aumento del 60% respecto al año anterior. Aunque AWS mostró un backlog que creció 40% interanual, Thompson cuestiona cuánta de esa demanda proviene de terceros vs. uso interno de Amazon. La pregunta central: ¿estamos ante inversión visionaria o sobreinversión en un mercado donde la IA se está comoditizando?

Thompson contrasta su postura sobre Amazon (nervioso) con su visión más optimista sobre Google. La diferencia clave: Google monetiza IA directamente en búsqueda y publicidad, mientras Amazon está apostando masivamente en infraestructura cloud donde los márgenes podrían comprimirse si los modelos de IA se comoditizan. Las preocupaciones incluyen márgenes más bajos, quema de efectivo acelerada y emisión de deuda para financiar el CapEx.

El artículo también explora el concepto de "Commodity AI" — la posibilidad de que los modelos foundation se conviertan en commodities, lo que beneficiaría a quienes los usan (aplicaciones) pero presionaría a quienes invierten en infraestructura sin diferenciación clara. Para inversionistas, la pregunta es si AWS puede mantener su pricing power o si la competencia de Azure y GCP erosionará márgenes en un mundo donde el modelo subyacente importa menos.

Key Takeaways:

• $200B CapEx — Amazon planea gastar $200B en 2026, 60% más que el año anterior, principalmente en infraestructura de IA/cloud
• AWS backlog +40% YoY — Señal positiva de demanda, pero incertidumbre sobre cuánto es interno vs. terceros
• Commodity AI thesis — Si los modelos se comoditizan, invertir masivamente en infra cloud es riesgoso
• Google vs Amazon — Thompson más bullish en Google (monetización directa via ads) que en Amazon (apuesta de infraestructura)
• Cash burn & deuda — Márgenes bajo presión, Amazon emitiendo deuda para financiar expansión

Notes: Artículo relevante para thesis de inversión en hyperscalers. Complementa el framework de "Endgame Positions" — Amazon tiene operational advantage (#9) pero la pregunta es si infrastructure alone es suficiente moat.

Arthur Brooks: Suffer Like a Marine

Author: Arthur Brooks | Source: The Free Press | Date Added: 2026-02-06 Tags: philosophy, happiness, stoicism, mental-health, dalai-lama

Summary: Arthur Brooks explores the Buddhist/Tibetan concept of managing pain through lowering resistance rather than eliminating pain. Based on his work with the Dalai Lama.

Key Takeaways:

• Suffering = Pain × Resistance — You can't eliminate pain, but you can lower your resistance to it
• Resistance means indulging in fear, anger, or self-pity when feeling pain
• Nonresistance means accepting pain without those emotions — the pain just is
• His Marine daughter experiences high pain but low suffering because she doesn't resist
• One method: morning prayer asking for whatever the day brings, including the hard parts

Notes: Mentions his new book "The Meaning of Your Life" with Dalai Lama endorsement.

Del Arco Norte a Nordstrom: Entrevista con el CEO de Liverpool

Author: René Lankenau H | Source: Whitepaper.mx | Date Added: 2026-02-06 Tags: retail, mexico, liverpool, nordstrom, fintech, ecommerce

Summary: Entrevista exclusiva con Enrique Güijosa, CEO de Liverpool, sobre la importancia del negocio financiero, evolución hacia comercio unificado, y la inversión en Nordstrom.

Key Takeaways:

• Negocio financiero — 11% de ingresos pero >50% de utilidad de operación. 8.5M tarjetas, tercer emisor más grande de MX. Cartera ~$80,000M MXN.
• Tarjeta cerrada — 80% del portafolio. 50%+ de ventas Liverpool se pagan con ella. En digital, 77% de ventas con tarjetas propias.
• Comercio unificado — De 9% digital en 2019 a 33% hoy. 34M paquetes/año. 90%+ del fulfillment es desde tiendas.
• Arco Norte — Inversión ~$1B USD en hub logístico automatizado en Edo Mex.
• BYD — Salieron del negocio automotriz. "Zapatero a tus zapatos" — tasas de crédito automotriz más bajas que tarjeta.
• Nordstrom — Liverpool tiene 49.9%. Invirtieron $1.2B USD. Ya les pagaron los $400M del préstamo. "Ha ido bien."
• Aprendizajes de Nordstrom — Analytics de clientes, e-commerce (40%), presentación de catálogo, programa de lealtad.

Notes: Güijosa fue CFO de P&G Chile/Brasil, Palacio de Hierro, y Liverpool por 13+ años antes de ser CEO en abril 2024.

Benedict Evans: AI and Software (Stratechery Interview)

Author: Benedict Evans & Ben Thompson | Source: Stratechery | Date Added: 2026-02-05 Tags: ai, software, enterprise, startups, tech-analysis

Summary: Annual interview where Evans and Thompson discuss whether AI is killing software companies, what role software plays in enterprises, OpenAI's search for a moat, and future of devices.

Key Takeaways:

• Typical big US company has 350-450 separate SaaS apps
• Making software is now 10-100x faster/cheaper with AI — like AWS did for infrastructure
• Software = stored processes, someone deciding "this is the right way to do this"
• The middle-manager-builds-tools thesis is "delusional"
• Real question: How far up the stack do foundation models go? Does your product become just a ChatGPT feature?
• SaaS walls are collapsing — competition from horizontal expansion is increasing

Notes: Part of annual Evans-Thompson tradition. Luis asked for summary via email.

Welcome to February 7, 2026: The Singularity Update

Author: Dr. Alex Wissner-Gross (@alexwg) | Source: X/Twitter | Date Added: 2026-02-07 Tags: ai, singularity, tech-trends, robotics, semiconductors, crypto, space

Summary: Weekly roundup of tech developments signaling the approaching singularity. Covers AI workforce automation, model competition, cryopreservation breakthroughs, and space compute.

Key Takeaways:

• AI as Workforce — OpenClaw agents running 24/7 as "synthetic employees" in China. OpenAI requiring all employees to code via agents by March 31.
• Model Race — Claude Opus 4.6 now #1 on Vals Index and Code Arena. Grok 4.20 dominates finance with 34% return in Alpha Arena trading sim.
• Semiconductors Exploding — Memory prices up 80-90% in Q1. Global chip sales projected to hit $1T this year.
• Cryopreservation Breakthrough — 21st Century Medicine achieved perfect ultrastructural preservation of rabbit brain via vitrification. First proof of feasibility for human cryopreservation.
• Space Compute — NY lawmakers introduced data center moratorium. SpaceX shifting focus to lunar disassembly for AI data centers, Mars missions delayed to March 2027.
• Robotics — Waymo using "Waymo World Model" (Genie 3) for training. Tesla FSD reportedly saving heart attack victims by driving to hospitals faster than ambulances.
• Economy — Palmer Luckey's Erebor Bank got national charter for 24/7 crypto banking. AI.com sold for $70M.

Notes: Weekly singularity digest format. Good pulse check on frontier AI developments.

End Game Positions: What to Own When AI is Abundant

Author: Michael Bloch (@michaelxbloch) | Source: X/Twitter | Date Added: 2026-02-08 Tags: ai, investing, endgame, thesis, portfolio-strategy

Summary: Framework for identifying what assets become MORE valuable as AI intelligence becomes abundant and cheap. Based on Will Manidis's "End Game Play" concept — reason backward from the terminal state.

Key Takeaways — The 12 Endgame Positions:

1. ⚡ Energy — AI needs power. Can't think your way to more electrons. (NextEra, Oklo, Base Power)
2. 🔩 Atoms — Can't download steel. Physical matter requires physical processes. (TSMC, ASML, Nvidia, Tesla)
3. 💰 Capital — Physical things cost real money. Chip fab = $20B, nuclear plant = $10B. (JPM, BlackRock, Goldman)
4. 📜 Regulatory Permission — Governments move at speed of politics, not tech. License to operate = durable moat. (SpaceX, Visa, Stripe)
5. ⚖️ Trust & Accountability — When AI can do anything, who's responsible? Human liability persists. (Licensed professions)
6. 📊 Proprietary Data — Unique datasets accumulated over decades. When all models have same architecture, data differentiates. (Bloomberg, Palantir, Epic)
7. 👁️ Human Attention — Finite resource. When content/outreach cost → 0, filtering becomes bottleneck. (Platforms that earn focus)
8. 🔗 Network Effects & Liquidity — Can't manufacture demand with intelligence. Critical mass takes time. (Uber, Airbnb, DoorDash)
9. 🏭 Operational Advantage — Intelligence replicable, installed base is not. Driver networks, merchant relationships, decade of learnings. (Amazon, Walmart, FedEx)
10. 🛡️ Security — More capable AI = more capable attacks. Offense & defense both scale. (CrowdStrike, Palo Alto)
11. 🌍 Physical Space — Can't create more land near substations, orbital slots, or spectrum. (Owners of AI infra locations)
12. 🧠 Intelligence Itself — Frontier model training requires billions + scarcest talent. Few companies can operate at scale. (xAI, OpenAI, Anthropic, Google, Meta)

Core Question: Will this asset get stronger or weaker as intelligence gets cheaper? If stronger → playing for endgame.

Notes: Excellent framework for evaluating 2026 portfolio positioning. Compare against current holdings.

Is This Something Big? — The Weekend Reader

• Date: 2026-02-15
• Source: The Weekend Reader (Substack)
• Topic: AI disruption, future of work
• Summary: Analiza el ensayo viral de Matt Shumer (80M+ views) sobre IA cruzando un umbral en 2026. Tres perspectivas: Shumer (alarmista — todo knowledge work en riesgo, GPT-5.3 se ayudó a escribir a sí mismo), escépticos (Doorman Fallacy — el valor invisible del trabajo humano), e histórica/teológica (la máquina de tejer no destruyó Inglaterra, la hizo rica). Consenso: aprende IA ahora.
• Key quote: "The economy is not a fixed pie. It's a garden. And technology is rain."
• Tags: AI, future-of-work, investing, theology