Erik Torenberg

Ben Horowitz × Martin Casado × Raghu Raghuram × Erik Torenberg

a16z’s new AI infrastructure fund captures a founder migration into hardware, with top-founder hardware pitches rising from roughly 3–5% to “north of 20% or 30%.”Hyperscaler capex, booked-out GPUs and resale premiums support opportunities across chips, power and cooling, while grid shortages, regulation and uncontrolled agent spending remain constraints.

Erik Torenberg × Nathan Labenz × Kyle Corbitt

RL can push capable open-weight models higher by reinforcing rare outcome-changing decisions while preserving pretrained “grooves,” unlike SFT’s broader risk of catastrophic forgetting.For enterprise agents, the clearest wedge is latency: RL can bring small models to roughly 30% of frontier latency and improve cost per token by at least an order of magnitude, though slower iteration and reward hacking remain risks.

Marc Andreessen × Erik Torenberg

Andreessen sees AI three years into an 80-year revolution, with customer revenue validating unusually rapid adoption across an already-built global distribution network.Falling token costs, competing chips, smaller models, and US-China rivalry could expand demand dramatically, while state regulation remains a major unresolved risk.

Erik Torenberg × Nathan Labenz × Zvi Moshowitz × Greg × Eugenia Kuyda × Ali Behrouz × Logan Kirkpatrick × Jungwon Hwang

AI’s strongest proof point is its performance alongside Nathan Labenz’s son’s oncologists, with minimal residual disease below one cell per million after remission before round two.Claude Opus 4.5 may be software AGI, but jagged failures and holiday hype leave full AGI unresolved, while context management, multi-model judgment, and infrastructure financing remain key risks.

Erik Torenberg × Patrick Hsu × Jorge Conde

Arc is combining neuroscience, immunology, machine learning, chemical biology, and genomics to reduce organizational and experimental latency, with virtual cells aimed at navigating diseased cells toward healthy states.The commercial prize depends on better targets, drug design, higher clinical success, lower capital intensity, and shorter timelines, but physical testing and human endpoints remain serial bottlenecks while virtual-cell capability sits somewhere between GPT-1 and GPT-2.

Erik Torenberg × Mark Cuban

Cuban’s clearest business thesis pairs transparent pricing and AI-native operators with broad employee equity, arguing that ESOP workers earn dramatically more and turn over less while small firms gain immediate productivity tools.He would replace opaque insurers and PBMs with patient shopping, catastrophic reinsurance, and government-guaranteed financing, while NBA media rights and second-apron constraints make subscription growth and patient team-building the next tests.

Erik Torenberg × Martín Casado × Leo Polovets

Martín Casado argues that ignoring consensus is dangerous because early venture markets are “pretty darn efficient,” while a startup dependent on follow-on capital must eventually become fundable.The highest-alpha bets may begin outside consensus and then reprice sharply, but rapid markups can create fragility; the unresolved test is whether high-priced winners actually produce superior venture returns or merely reflect price arbitrage.

Erik Torenberg × Steven Sinofsky × Balaji Srinivasan

Blocking Big Tech exits can starve startups of capital and strengthen incumbents: DOJ intervention in JetBlue’s acquisition of Spirit was followed by Spirit going bust, while acquisitions fund challengers through incumbent “surrenders.”AI’s platform shift is driving faster acqui-fires, including Google’s Windsurf deal, while copyright litigation, energy constraints and restrictions on Chinese models could squeeze US leadership.

Erik Torenberg × Bryan Kim × Roy Lee

Cluely is testing whether distribution can discover product-market fit faster than conventional development, using more than 1 billion views and sales-call videos that Roy says generated over $1 million in enterprise revenue.Its translucent screen-and-audio overlay targets an emerging AI interface, while follower-based hiring and paid creator production industrialize acquisition; the unresolved risk is that technically simple features are copied before the land grab becomes durable.

Erik Torenberg × Nathan Labenz × Matt Perault

a16z favors policing harmful AI uses while keeping research broadly open, arguing that compute or training-cost thresholds could reinforce concentration among only three to 10 companies.The unresolved risk is whether ex-post enforcement can address irreversible harms, while temporary rules, liability regimes, and the Trump administration’s national AI action plan in June or July shape market access.