Source List
Working source map for Where Value Lands. Organized by five thematic areas that map to the paper’s chapters. Key media channels are listed separately at the bottom — mine them for episodes, don’t treat them as standalone sources.
Note
This is a prioritized shortlist, not an exhaustive bibliography. Goal: enough coverage to calibrate before systematic research begins. See 02_deep-research-prompts for the agent-ready research briefs.
1. Vision & AI Trajectory
Maps to: Section 2 — The Stack
Thinkers who frame what AI is becoming at the highest level of abstraction. Primary input for the reference frame the paper builds on.
| Source | Key Material | Why |
|---|---|---|
| Andrej Karpathy | ”Software 3.0” essay/talk; “LLM OS”; YC talk; Eureka Labs interviews | Best conceptual map of the stack from a practitioner-builder; defines the L0–L7 terrain |
| Dario Amodei | Machines of Loving Grace; Dwarkesh + Lex Fridman interviews | Canonical optimistic long-run framing; calibrates upside scenarios |
| Sam Altman | Stratechery, All-In, Bloomberg interviews (recent) | Platform-level vision from the lab with the largest deployment surface |
| Demis Hassabis | Lex Fridman, Hard Fork, 60 Minutes | Scientific / AGI framing; DeepMind approach as counterpoint to OpenAI |
| Mustafa Suleiman | The Coming Wave talks; Microsoft AI interviews | Consumer-AI + safety framing; useful foil for the infrastructure-heavy view |
| Ilya Sutskever | Post-SSI public material (limited — curate what exists) | Frontier-capabilities view; safety-first lens |
2. Economics, Strategy & Value Distribution
Maps to: Sections 3 + 6 — Where the Margin Lands / Who Wins
Platform economics, capex, infrastructure dynamics, and the U-curve analogies. The closest the secondary literature comes to the paper’s central thesis.
| Source | Key Material | Why |
|---|---|---|
| Ben Thompson | Stratechery — Aggregation Theory applied to AI; AI platform and distribution pieces | Direct theoretical backbone for the distribution-layer hypothesis |
| Brad Gerstner | BG2 podcast (with Bill Gurley) — AI capex and economics episodes | Clearest investor framing of the infrastructure build-out and its economics |
| Bill Gurley | BG2 + platform-transition interviews | Supply- vs. demand-side moats; platform-shift historical patterns |
| Martin Casado | a16z talks on infrastructure economics | Infrastructure-layer economics; bottom of the U |
| Dylan Patel | SemiAnalysis — GPU economics, datacenter, supply chain | Best living source on the actual economics of compute; highly quantitative |
| Jensen Huang | Acquired podcast (2023, ~3 hrs); keynotes | NVIDIA’s position at the infrastructure layer; first-hand strategic view |
| Elad Gil | Podcast interviews, especially on agents and the application layer | Investor-practitioner lens on where the application layer stands |
| Packy McCormick | Not Boring strategic essays + YouTube | Long-form strategic synthesis; useful for narrative texture |
| Chamath / All-In Pod | AI capex and bubble-debate episodes | Stress-tests the infrastructure bull case; useful bear-side data point |
3. Technical Frontier: Agents, Models & Open Source
Maps to: Sections 2 + 4 — The Stack / Scenarios
Coding agents, orchestration, open-weight models, and edge computing. The empirical reality of what the middle of the stack actually looks like in practice.
| Source | Key Material | Why |
|---|---|---|
| Aravind Srinivas | Lex Fridman, Dwarkesh interviews | Perplexity as a live experiment in distribution-layer value capture |
| Varun Mohan / Cognition | Devin interviews (recent) | Coding-agent frontier; L4–L5 of the stack in practice |
| Anysphere / Cursor team | Available interviews | IDE-layer as distribution moat; strong U-curve relevance |
| Harrison Chase | LangChain talks on orchestration | Orchestration-layer nuance; useful even as a critical / skeptical data point |
| Yann LeCun | World models + open-source architecture talks | The alternative architectural thesis; structural counterpoint to the transformer-scaling camp |
| Arthur Mensch | Mistral interviews | Open-weight economics from the inside |
| DeepSeek / Liang Wenfeng | Lex × Dylan Patel + Nathan Lambert episode; secondary coverage | Chinese frontier model dynamics; open weights at scale; geopolitics of commoditization |
| Nathan Lambert | Interconnects newsletter + podcast | Small model + open-source lens; rigorous and independent |
4. Skeptics & Bear Case
Maps to: Section 4 — Scenarios (stress-tests all hypotheses)
Counter-narratives, ROI critiques, structural objections. Non-negotiable for a credible paper — if these aren’t engaged, the paper reads as a manifesto.
| Source | Key Material | Why | Status |
|---|---|---|---|
| Gary Marcus | One or two rigorous pieces (diminishing marginal return beyond that) | Capability-overhype corrective; sets a ceiling on claims | Stub via synthesis — marcus-world-models-failure |
| Ed Zitron | Better Offline podcast | Revenue and deployment reality check; product lens | Stub via synthesis — zitron-circular-economics |
| Goldman Sachs | Gen AI: Too Much Spend, Too Little Benefit? report + coverage | Institutional ROI scepticism; quantitative; citable | Stub via synthesis — goldman-sachs-too-much-spend |
| Daron Acemoglu | AI economics interviews and papers | Labor-economics and productivity challenge; academic rigour; hardest to dismiss | Stub via synthesis — acemoglu-simple-macroeconomics |
Note
All four sources are currently summarized only through the internal synthesis bear-case-synthesis. Direct ingestion of each primary source is the next priority for this section to be citation-ready.
5. Geopolitics & Governance
Maps to: Section 5 — Geographic Split
US/China race, EU regulation, sovereign AI dynamics. The question — whether the AI value stack has one global shape or three different optima — is now answered (provisionally: three optima) by the internal synthesis geopolitics-global-ai-divide and the concept page divergent-value-stack-optima. Direct ingestion of each primary source below is the next priority for citation-readiness.
| Source | Key Material | Why | Status |
|---|---|---|---|
| Anu Bradford | Digital Empires; Tech Policy Press interview | EU regulatory architecture; tripartite “digital empires” frame | Stub via synthesis — bradford-digital-empires |
| Jeffrey Ding | Running the Right AI Race (RAND, 2025); China’s Diffusion Deficit (FPRI, 2024); ChinAI newsletter | GPT diffusion-marathon thesis; structural reason for China’s downstream pivot | Stub via synthesis — ding-diffusion-marathon |
| Kai-Fu Lee | Capgemini interview + FT coverage (2024); 01.AI B2B pivot | First-hand strategic view of China’s bottom-heavy optimum | Stub via synthesis — lee-01ai-pivot |
| Helen Toner | Lawfare Daily / 80000hours podcast (Sept 2024) | “Steerability” + “adaptation buffers” critique of US non-proliferation framing | Folded into geopolitics-global-ai-divide |
| DeepSeek coverage | Economic Times, Jan 2025 + Lex × Dylan Patel × Nathan Lambert | ”Sputnik moment” empirical anchor for chip-embargo blowback | Folded into open-weight-asymmetry |
| EU AI Act backlash | FintechWeekly (July 2025) — open letter from European founders/VCs | Direct market evidence that AI Act is currently a handicap | Folded into brussels-effect-and-mirage |
| Brussels Mirage | LSE Law Review, April 2026 — predictive-policing case study | Structural critique of Brussels Effect in AI: form travels, substance doesn’t | Folded into brussels-effect-and-mirage |
| Sovereign AI coverage | WEF (Nov 2024) — Denmark, Italy, Sweden case studies | European hardware play; “sovereignty traps” | Folded into sovereign-ai |
| Global AI Index (Tortoise) | EUTechloop / WEF coverage (2024–2025) | Quantitative baseline: US/China dominate; EU populates secondary tier | Folded into geopolitics-global-ai-divide |
| Sifted | European AI sovereignty coverage | On-the-ground EU narrative; scenario colour | Not yet ingested |
6. SME Diffusion & Policy
Maps to: Sections 4 + 6 — Scenarios / Who Wins (demand-side counterpart to the supply-side stack)
The diffusion-side question: who actually adopts AI in the long tail of the real economy, and what determines whether they can? Bears directly on the bottom-right of H2_u-curve-of-value and operationalises diffusion-vs-innovation.
| Source | Key Material | Why | Status |
|---|---|---|---|
| OECD (2025) | AI adoption by small and medium-sized enterprises — OECD discussion paper for the G7 | SME-facing demand-side counterpart to the supply-side L0–L7 stack: four-enabler quartet (connectivity, AI-enabling inputs, skills, finance), four-stage adopter taxonomy (Novices → Optimisers → Explorers → Champions), G7 country profiles, ~70-country comparator base, OECD policy recommendations. Load-bearing for diffusion-vs-innovation, H2_u-curve-of-value bottom-tail, divergent-value-stack-optima, fusion-skills, sovereign-ai, foundational-enablers. | Ingested — oecd-sme-ai-adoption-2025 |
Key Media Channels
Not standalone sources — mine for specific episodes using 02_deep-research-prompts as a guide.
| Channel | Why |
|---|---|
| Dwarkesh Podcast | Highest signal-per-minute on this topic; all major lab founders have appeared |
| Acquired | Company-history depth; NVIDIA and TSMC episodes are mandatory |
| Lex Fridman | Long-form; AI lab founders; useful for full-picture calibration |
| No Priors (Sarah Guo / Elad Gil) | Application-layer investor lens |
| BG2 (Gerstner / Gurley) | Infrastructure and platform economics |
| Latent Space (swyx) | Technical operator lens; strong practitioner guest list |
| Hard Fork (NYT) | Mainstream reader calibration; keeps the paper honest on clarity |
Priority Queue
If time is limited before the research sprint begins:
- Karpathy — Software 3.0 / YC talk
- Dwarkesh × Dario Amodei (latest episode)
- Acquired × Jensen Huang (2023)
- BG2 — AI capex episode (Gerstner + Gurley)
- One bear-case piece (Zitron or Goldman) — so the paper doesn’t write itself as a manifesto