- Asymmetric Fragility: Why Vibe Coded Software Breaks Harder and Takes Longer To Recover↗ 15 Nov 2026
AI-assisted 'vibe coding' that prioritizes speed over comprehension creates asymmetric fragility: fast creation masks hidden complexity, leaving systems vulnerable to cascading failures and slow recovery—as evidenced by recent Microsoft 365 and FAA outages.
- Interview With Jason Ambrose, CEO of Backstory↗ 10 Nov 2026
Jason Ambrose, CEO of Backstory, discusses where AI agents are actually proving useful in enterprise sales—automating everything around the conversation while humans handle trust and judgment in live deals.
- The Game Theory Of Data Centers. Why The MidTerm Election Results Won’t Impact It At All.↗ 8 Nov 2026
Political opposition to data center construction won't meaningfully slow AI infrastructure buildout because zoning, interconnection, and utility decisions live at state and local levels where incentives still favor building. The real brake will be inference efficiency: token costs are collapsing 280x in two years, bending power demand from exponential growth into an S-curve.
- ChatGPT Is Stealing Questions From Google, But Not Money↗ 5 Nov 2026
ChatGPT captures a fifth to third of Google's search volume but sends nearly zero traffic downstream, while Google retains 85% of transaction-intent queries and 93% of navigation/local searches. The threat to Google is real but narrower than headlines suggest—LLMs dominate question-asking but not money-making queries.
- Proprietary Data Isn’t The Moat: My Interview with Todd Gottula From Clarify Health↗ 3 Nov 2026
Clarify Health CEO Todd Gottula challenges the common VC belief that proprietary data alone creates an AI moat, arguing the reality is more nuanced in healthcare AI.
- Silicon Valley Is Wrong About AI, And Wall Street Is Right↗ 1 Nov 2026
Silicon Valley's AGI bet frames AI as a zero-marginal-cost software monopoly, but Wall Street's capex-focused view is more accurate: AI is a capital-intensive utility with falling unit prices, commoditizing models, and value accruing to those controlling constraints and workflows, not frontier labs.
- Red Cat (RCAT) Lost a Drone Contest. It May Be Winning the Software War.↗ 27 Oct 2026
Red Cat's underperformance in a Pentagon drone competition obscures its real strategic asset: software for coordinating multiple drones from different vendors into a unified force—a capability aligned with the Pentagon's evolving AI warfare strategy.
- The AI Bubble Is Real, But Not Where You Are Looking↗ 25 Oct 2026
AI demand splits into five markets with vastly different economics: venture-funded startups and corporate POCs fund AI from balance sheets (bubble-prone), while scaled companies deploying AI for cost reduction fund from operating budgets (durable). The valuation risk hinges not on technology but on customer mix and revenue source.
- Who Owns What AI Makes? An Interview With DeepLaw Founders↗ 20 Oct 2026
Interview with DeepLaw founders on AI ownership, patent strategy, and legal implications—covering authorship law, inventor notebooks, and AI's impact on legal billing models.
- Beyond The GPU: Second Order Thinking In AI Infrastructure↗ 18 Oct 2026
Second-order thinking in AI infrastructure: as GPU deployment scales, the real constraints shift from chips to power generation, thermal management, optical networking, and edge inference. Infrastructure winners will be in cooling, energy, and interconnect—not foundation models.
- Arm Holdings (ARM): The Royalty Machine Meets the AI Data Center↗ 14 Oct 2026
Arm is transitioning from a pure IP licensor collecting royalties to a dual-engine business adding merchant silicon (AGI CPU). The shift could expand revenue tenfold but introduces manufacturing costs and channel conflict with customers who currently pay Arm royalties.
- AI Progress Is Stalling. Now Is The Best Time To Invest.↗ 11 Oct 2026
As AI frontier progress slows, the investment opportunity shifts from building foundation models to deploying existing AI across enterprise workflows. The author argues that legacy infrastructure, data integration, and adoption friction—not model capability—are the real bottlenecks, opening a decade of value capture in vertical AI, AI-enabled services, and proprietary data plays.
- AvePoint: The $15 Feature That Could Eat a $3B Company↗ 8 Oct 2026
AvePoint's stock has rallied 40% on an AI governance positioning, but Microsoft's $15 Agent 365 feature bundled into E7 may commoditize the category AvePoint is betting its future on—despite solid Q2 numbers, AgentPulse adoption hasn't yet moved net retention, and the company is pricing in revenue still in pipeline.
- In AI, Fast Means Fragile: How Wall Street Short-Termism Impacts AI Deployments↗ 4 Oct 2026
Wall Street's demand for immediate AI announcements is driving companies to ship commodity chatbots that erode margins and competitive advantage. True AI value requires patient, multi-year restructuring of data and workflows—a thesis the market hasn't yet priced in.
- Investing in AI Interviews: Ash Ashutosh, CEO at Pinecone↗ 29 Sept 2026
Interview with Pinecone CEO on the company's evolution from vector database to agentic knowledge platform and the future of agent knowledge management.
- Should Investors Be Concerned About The Recent AI Safety Incidents?↗ 27 Sept 2026
Recent AI safety incidents raise questions about existential risk, but the author argues doomsday scenarios rest on unvalidated assumptions—multiple competing AIs, controllable systems, and deployment harnesses suggest AI risk, while real, is overplayed; investors should not be overly concerned.
- The 2028 Handoff: Silicon Motion’s (SIMO) Bear Case and Bull Case Arrive in the Same Year↗ 24 Sept 2026
Silicon Motion, a memory controller chip maker riding AI-driven NAND shortages, faces a 2028 inflection when new fab capacity comes online—marking either a cyclical peak (bear case) or a transition to higher-margin controller-only businesses (bull case). The stock's 30% discount to highs reflects this uncertainty.
- The Owning Phase of AI Part 3: Companies To Research↗ 20 Sept 2026
Eight public companies positioned to build proprietary AI models as small models and open-source alternatives make internal ownership economically viable—watch S&P Global, Meta, Adobe, Thomson Reuters, Axon, Deere, Relx, and IQVIA for signs they're moving from API reliance to model ownership.
- Stress Testing Anthropic’s Revenue Pre-IPO↗ 17 Sept 2026
A stress-test analysis of Anthropic's revenue trajectory pre-IPO, arguing the company faces a duration-mismatch risk—heavy upfront infrastructure costs betting on $190-200B revenue by 2028—where a 30%+ revenue shortfall would force costly renegotiation rather than near-term solvency collapse.
- Investing in AI The Book, Is Live. Here Is Your Free Copy↗ 15 Sept 2026
Announcement of a new book on AI investing for general investors, with free access for newsletter subscribers and a request for reviews and media introductions.
- The Owning Phase of AI Part 2: When Do The Economics Make Sense↗ 13 Sept 2026
AI companies face a critical economics choice: renting API inference at scale erodes margins via super-linear token costs, while owning and distilling models into fine-tuned assets flips the structure to predictable, depreciating costs and defensible moats. The framework clarifies when ownership's upfront investment pays off.
- Nebius: A Software Company With A Data Center Attached↗ 10 Sept 2026
Nebius is shifting from a capital-intensive data center operator to an asset-light software platform company, partnering with external providers to finance infrastructure while Nebius focuses on software, customer relationships, and AI services—a model that could reshape its business model and valuation.
- Reader Submission: A Deeper Dive on Dell↗ 9 Sept 2026
Dell's AI server dominance generates revenue without proportional margin because it assembles around scarce accelerator silicon owned by others; the investment question is whether Dell can own the next binding constraint (power, cooling, or orchestration) before specialists claim it.
- The “Owning” Phase of AI Part 1: What Investors Should Look For In Companies Making Their Own Models.↗ 6 Sept 2026
Companies are shifting from renting frontier AI models to building their own; investors should identify targets with proprietary data, API-cost tipping points, high-stakes domains, workflow capture, and data sovereignty needs—not just model announcements.
- The AI Boom And Dell: Can The Company Capture More AI Value?↗ 3 Sept 2026
Dell is capturing AI server revenue at scale ($60B expected FY27), but most margin flows to GPU/chip makers; the real question is whether Dell can leverage these deployments into higher-margin services, storage, and financing to convert volume into durable profit.
- How To Think About The Negative ROI on AI Investment↗ 30 Aug 2026
Most AI investments show negative ROI today not because the technology is broken, but because pricing reflected a lottery-ticket bet on AGI rather than underwriting actual business use cases. As compute costs collapse via cheaper chips, smaller models, and better infrastructure, many 2025 failures will pencil out profitably by 2027 without changing anything but the bill.
- Flex Ltd: Will Power Be The New Silicon?↗ 26 Aug 2026
Flex Ltd. argues that AI infrastructure's bottleneck is shifting from silicon to power delivery—a contrarian bet that could reshape which vendors win in the AI buildout.
- The Speed of Thought: What If Intelligence Has a Universal Limit?↗ 23 Aug 2026
If intelligence has a physical ceiling like other universal quantities, and humans are already near it, then AGI promises are overblown—and hard problems like climate and aging remain ours to solve through policy and iteration, not superintelligence.
- My Brief Interview With Paddy Srinivasan, CEO of DigitalOcean, On Their AI Growth↗ 19 Aug 2026
A video interview series debut featuring DigitalOcean's CEO discussing the company's AI strategy, GPU infrastructure positioning, and how it competes against hyperscalers in mid-market inference and agentic workloads.
- Is There An Inference Glut Coming?↗ 16 Aug 2026
Analysis of whether AI inference demand will outpace datacenter supply through 2028. The author argues a glut is unlikely but possible, mapping supply/demand dynamics, efficiency gains, and the key monitorables that would signal a shift.