Food for Thought
A curated collection of articles, videos, and podcasts that have shaped how I think about investing, technology, and the world. Each entry includes a few things I took away from it.
Finding The 1% of Stocks That Matter | Henry Ellenbogen
- 1/ Durable Capital hunts the rare 1% of public companies (~40 stocks) capable of compounding wealth at 20% annually over a decade
- 2/ Founders on their second venture have a distinct edge — total clarity on solving edge cases, managing scale, and aligning their organization from day one
- 3/ AI as the new Kaizen: just as physical supply chains benefited from decades of cost deflation, AI represents continuous improvement applied to human processes and white-collar IP labor
- 4/ Only initiate an early-stage investment if actively excited about buying more shares at higher prices once the business proves its competitive advantage
- 5/ Writing detailed investment memos is essential for executing multi-year post-mortems and intellectual honesty
- 6/ The friction and scrutiny of public markets can force crucial discipline, efficiency, and cultural alignment during difficult corporate transitions
YouTube · Durable Capital
Dan Sundheim of D1 Capital on the Art of Public Market Investing
- 1/ D1 applies the exact same deep, fundamental research process and 3–5 year time horizon regardless of whether a company is public or private
- 2/ The danger of selling winners prematurely
- 3/ Short-squeezes (like GameStop) proved that risk management cannot be reactive; short books must be highly diversified and sized small enough to withstand irrational melt-ups
- “You’re always going to make the most money from multiple expansion… sometimes people are skeptical of a business model, and then over time it gets proven out that the business model doesn’t deserve to have that skepticism, and then the multiple expands.”
- 4/ D1 requires analysts to maintain a weekly “mock portfolio” that ties stated convictions directly to compensation
- 5/ Sundheim halted investments in China because the government’s heavy-handed control over resources and lack of due process create uninvestable uncertainty
YouTube · D1 Capital
The Benchmark Partnership
- 1/ Maximizing happiness over AUM scale
- 2/ The firm thrives as an equal partnership because its founding members intentionally gave up their legacy and residual brand economics to empower the next generation — essentially “refounded” with each new partner
- 3/ Board role viewed not as governance but as a trusted co-founder who actively spars with entrepreneurs to sharpen their thinking; firms pitching “no board seat” fundamentally misunderstand this
- 4/ Distinguishing rare “entrepreneurs” who inherently challenge systems from “founders” who are riding a temporarily attractive trend
- 5/ AI investments succeeded not through a top-down thematic market map, but through relentless focus on backing incredibly special, non-consensus founders
YouTube · Benchmark
Ilya Sutskever — From the Age of Scaling to the Age of Research
- The era where scaling (adding compute and data) was the primary driver of progress is concluding as pre-training data becomes finite; breakthroughs now depend on novel algorithmic ideas
- AI models ace difficult benchmarks but fail basic real-world logic (e.g., oscillating between bugs in code), highlighting a deficit in reliable generalization vs. humans
- Humans require drastically fewer samples to learn (e.g., 10 hours for driving vs. millions of simulations for AI), proving more efficient ML principles exist — Moravec’s paradox
- SSI prioritizes a focused research path to superintelligence, insulating itself from product-cycle pressures to solve fundamental bottlenecks like sample efficiency
YouTube · SSI
Sir Demis Hassabis on The Future of Knowledge
- Games (chess, Go) are not ends but data-rich, rule-governed testbeds for developing general learning algorithms
- DeepMind’s ultimate goal: use AI to solve “root node” problems like protein structure (AlphaFold) to unlock entire fields of medicine and material science
- Any pattern in nature can potentially be modeled by classical learning algorithms, challenging the necessity of quantum computing for biological simulation
- Future progress relies on “Astra” and world models that understand intuitive physics and spatial context beyond what text-based LLMs can capture
- AI development requires new global institutions (a “Technical UN” or “AI IAEA”) to monitor risks and ensure safe progress
YouTube · Institute for Advanced Study
Fei-Fei Li: Spatial Intelligence is the Next Frontier in AI
- Spatial intelligence — the ability to perceive, reason about, and interact with the 3D world — is the essential missing piece of AGI
- Vision took 540 million years to evolve, language less than one million, highlighting the vast complexity gap between LLMs and “World Models”
- ImageNet (2009) was a bold bet that a paradigm shift toward data-driven methods was necessary for machines to generalize
- World Labs aims to solve “delusional” problems in 3D world modeling, combining generation with physical reconstruction and causal reasoning
- Academia is now severely under-resourced compared to industry, requiring researchers to pivot toward interdisciplinary and theoretical North Stars
YouTube · World Labs
A War of Attrition — Missiles, Oil, and Global Power | The Loonie Hour
- Oil and gas are the new foundation of US military strength
- The current Middle East conflict is a war of attrition and depletion of US military inventories — the US is dangerously short on air defense missiles
- China and Russia benefit from a prolonged Middle East war; China has prepared by building massive refining capacity and stockpiling
- Global oil markets are no longer a free, frictionless trade
YouTube · The Loonie Hour
Future Forward: What India’s Economy Looks Like in 2026 | Open Dialogue
- India achieved growth despite massive fiscal and monetary headwinds
- Fiscal consolidation will slow in FY26; credit growth is reviving, state-level labor reforms are unlocking potential, and the real estate cycle is picking back up
- The Rupee’s depreciation is a capital account issue, not a fundamental problem — gold speculation effectively shorts the Rupee; the Rupee remains competitive on a trade-weighted basis
YouTube · Open Dialogue
Russell Napier on The Solid Ground, Anatomy of a Bear
- Economic theory relies on the “rational economic man,” completely ignoring the realities of psychology, sociology, and political behavior
- Financial repression forces savers to pay for government debt; the state uses regulations (like mandating pension fund purchases) to keep bond yields below inflation
- The US is adopting China’s style of “social capitalism” — abandoning pure free-market principles to heavily subsidize its own industries to compete with state-sponsored overcapacity
- Gold is critical because it lacks counterparty risk; value stocks will thrive in this new environment
YouTube · The Solid Ground
WW3, Energy, Markets & Politics with Doomberg
- The world is actively fighting an economic World War III; the Middle East is exclusively about oil and gas — the US no longer needs it for itself but uses its presence to constrain China
- Social media training algorithms show opposing sides starkly different, highly propagandized versions of reality — dangerous “parallel universes”
- Taiwan can be defeated without an invasion: complete reliance on imported LNG for power means China only needs a blockade to force capitulation
- Redeploying US defenses (Patriot and THAAD) from South Korea to the Middle East leaves Asian allies vulnerable to aggressive moves by North Korea and China
YouTube · Doomberg
Ben Thompson (Stratechery) on AI Ads, the End of SaaS, and the Future of Media
- “Agent commerce” could trigger brutal perfect competition — AI making optimal purchasing decisions strips emotion from commerce, devastating retail margins
- SaaS valuations are compressing as AI agents and self-serve tech limit headcount growth, forcing legacy businesses into massive valuation haircuts
- A major chip shortage is looming by 2029 — TSMC is underbuilding because the downside of excess fabrication capacity is too severe
- Google’s lack of optimization is its hidden strength — excess cash flow and organizational slack give it the flexibility to survive major disruption
Cheeky Pint (Stripe) · John Collison
Demis Hassabis & Dario Amodei — The Day After AGI (Davos 2026)
- AI may fully automate software engineering within 6–12 months — models could handle the entirety of what SWEs do end-to-end
- Once AI can write code and conduct AI research, it will dramatically accelerate next-generation models — closing the loop triggers exponential progress
- AI revenue scaling exponentially: Anthropic went $0 → $100M (2023) → $1B (2024) → $10B (2025)
- Geopolitics prevents a coordinated AI slowdown — chip export restrictions are the most critical lever for buying time
World Economic Forum · Davos 2026
Gokul Rajaram — Lessons from Investing in 700 Companies
- Human judgment is the most future-proof skill — with AI generating massive output, humans must determine what actually matters
- AI startups must own a scarce asset or control point to survive — proprietary data, financial movement, or hardware
- Non-deterministic software forces heavy evaluation (evals) — a PM’s primary role is now building evaluations to confirm if AI output is safe to ship
- The future of management is orchestrating fleets of AI agents rather than managing human employees
Invest Like the Best · Patrick O’Shaughnessy
Venky Ganesan (Menlo Ventures) — Inside Silicon Valley’s AI Gold Rush
- AI is the biggest platform shift of our lifetime — bigger than the internet, and the value will eventually move to the application layer
- Infrastructure is overvalued, applications are undervalued — too much capital going to AI infra, leaving apps as the opportunity
- India faces critical choices on sovereign AI — build closed-source, embrace open-source, or partner with US companies for localized models
- Tech shifts do not cause long-term job destruction — historical precedents (agriculture, ATMs) show AI will alter roles, not eliminate employment
YouTube · Menlo Ventures
Bret Taylor — Is AI Killing Software?
- “Vibe coding” will become the standard expectation — the idea that your software is something you can change yourself will be something we expect
- Software form factors will shift from browsers and forms to autonomous agents operating against databases
- Personal AI agents will become the internet’s front door — consumers will delegate browsing and purchasing to their agents
- Business model transitions (e.g., outcomes-based pricing) will be far harder for incumbents than adopting the underlying technology
Big Technology Podcast · Alex Kantrowitz
Sridhar Ramaswamy (Snowflake) — Data’s Role in the AI Era
- Enterprise AI demands a single source of truth — built on security, auditability, and trust; customer data is never used to train external models
- AI delivers extreme accelerations — software engineering projects that took four weeks now take forty minutes (100x speedup)
- Agile organizational setups matter more than pure hardware — team structure for fast deployment beats massive hardware spending
- AI is redefining corporate workflows — enterprises are using AI tools to completely replace legacy software and rip out old processes
Sequoia Capital · Training Data
Content coming soon.
Content coming soon.