The day in one read

The central tension defining the current business landscape is the collision between the rapid commoditization of intelligence and the persistent scarcity of human attention and trust. As artificial intelligence transitions from a novel tool to a foundational utility, companies are restructuring their operational models to leverage this abundance, while investors and strategists debate whether this shift signals a new era of economic expansion or a dangerous deflation of human value. The day’s developments reveal a market bifurcating into two distinct camps: those who view AI as a lever for aggressive growth and capital efficiency, and those who warn that the underlying assumptions of "cognitive scarcity" are obsolete. This divergence is visible in the aggressive debt strategies of private equity firms, the pivot toward performance-based marketing in B2B SaaS, and the emerging infrastructure required to support autonomous agents.

The Economics of Cognitive Abundance

A significant intellectual challenge to prevailing AI anxiety has emerged, arguing that the current panic is rooted in an outdated economic framework. Venkatesh Rao contends that contemporary fears of AI constitute "cognitive mercantilism," a worldview that treats intelligence as a finite, zero-sum resource to be hoarded by elites. Drawing on historical analogies from Joyce Appleby’s The Relentless Revolution, A History of Capitalism, Rao argues that capitalism itself emerged when society abandoned the mercantilist view of fixed wealth in favor of recognizing that wealth could be created and that ordinary people were sources of demand. Rao posits that AI will trigger a similar paradigm shift, transforming intelligence from a scarce production input into an abundant consumer good.

Rao identifies four core mercantilist beliefs that he argues are now obsolete: that intelligence is fixed, that the masses cannot be trusted with it, that supply control is paramount, and that most humans are intellectually useless. He suggests that just as agricultural improvements freed 80% of the population from farming to become consumers, AI will free most humans from the necessity of producing intelligence, allowing them to become significant consumers of it. This shift unlocks "demand elasticity" across six dimensions: price, access, complexity, continuity, personalization, and relational. Rao contrasts this optimistic view with "AI Malthusianism," which fears that an intelligence surplus will render humans redundant. Instead, he predicts a "cognitive industrious revolution," where individuals voluntarily increase cognitive effort because the returns on attention rise, mirroring how early modern workers increased labor days to afford new consumer goods. The political implication, Rao argues, is a shift from preserving cognitive jobs to securing a fair share of expanding cognitive consumption opportunities.

Infrastructure for Autonomous Agents

While the philosophical debate continues, the technical infrastructure required to support AI agents is undergoing a rapid maturation. Sahil Jain, CTO of U.com, argues that the current bottleneck in AI development is not model intelligence but the lack of supporting infrastructure, a situation he likens to self-driving cars lacking roads. Jain distinguishes between "core intelligence," provided by foundation models, and "augmented intelligence," which relies on external tools and services. He asserts that models will always require external tools, such as search, to access real-time information beyond their training cutoffs. As frontier labs like OpenAI and Anthropic verticalize to own the end-to-end stack, independent infrastructure companies must position themselves as neutral layers that co-evolve with model providers.

The operational requirements for agent search differ fundamentally from human search. Agents can execute 1,000 to 10,000 queries per second, a throughput no human can match. Unlike humans, who rely on precision and top-ranked results, agents prioritize recall over precision, consuming more context and tolerating varied latency. Jain notes that static public benchmarks like BrowseComp are insufficient because they allow for memorization and do not reflect production reliability. To address this, U.com is developing dynamic benchmarks like LiveQA and VerticalRTK to measure freshness and real-world performance. The company’s strategy relies on cross-model routing and post-training improvements to maintain value against verticalizing frontier labs, with customer success at scale—measured in tens of millions of monthly queries—serving as the ultimate validation of the infrastructure.

Performance-Based Growth in B2B SaaS

In the commercial sector, B2B SaaS companies are increasingly abandoning traditional flat-fee marketing models in favor of performance-based user-generated content (UGC). Whisperflow, which saw revenue grow 30X in the past year, exemplifies this shift. Hector Serrano, who leads the company’s paid UGC program, oversees a team of 50 creators launching 500 to 600 ads monthly. The strategy employs the Tribe platform, where creators are compensated via a percentage of ad spend or GMV rather than flat fees, with top creators earning up to $13,000 monthly. To mitigate early earnings volatility, Whisperflow implements a "smart retainer" model, guaranteeing a base payment that is offset by commission earnings if they exceed the threshold.

Serrano argues that B2B companies should treat enterprise buyers as individuals, using UGC to drive bottom-up adoption before closing enterprise deals. He emphasizes hiring young talent with proven social media growth over traditional credentials, noting that older founders often lack intuition for current digital channels. Serrano also highlights founder-led content as a long-term growth lever, citing Whisperflow’s YouTube channel, which reached 20,000 subscribers in three months. Crucially, Serrano advises against using AI for final creative assets, arguing that human-made content cuts through market noise more effectively than "AI slop." He notes that the team must remain experimental to avoid ad fatigue, as Meta’s Andromeda system groups similar ads, causing rapid performance decay. The team structure remains lean, consisting of a head of content, a media buyer, and specialized editors, reflecting a broader trend toward agile, high-signal marketing operations.

Debt Leverage and AI-Enhanced Moats

Private equity firms are leveraging AI to enhance the durability of their portfolio companies’ moats while using debt to compound free cash flow. Vending Spoons, co-founded by Luca Ferrari, operates a buy-and-improve model, having acquired approximately 50 to 60 companies including Airtable, AOL, Vimeo, Eventbrite, Komoot, and WeTransfer. Ferrari prioritizes long-term free cash flow compounding over short-term valuation, leveraging debt to fund acquisitions. The company maintains a leverage ratio of approximately 2.4x to 2.5x, with a self-imposed limit below 3x. To mitigate risk, Vending Spoons refinances most debt annually to keep maturities five years or more out and hedges 100% of debt against interest rate fluctuations. Ferrari asserts that the current risk of default is close to zero.

Ferrari argues that SaaS moats remain durable due to first-mover advantage, network effects, and proprietary data, rather than code complexity. He notes that enterprise software spend represents 8% to 10% of total corporate spend, making AI-driven replication of payroll or core software a low-upside proposition. Internally, AI handles 94% of coding, up from 0% two years ago, and at least 90% of data analysis tasks. Customer support is 100% handled by AI or external vendors, with human managers focusing on process architecture. The company has built over 50 in-house tools, including an AI agent named "Old ALT" that operates within Slack to execute tasks like code fixes and data retrieval. Ferrari’s approach suggests that AI is not destroying established SaaS moats but rather enhancing them by reducing the cost of maintenance and increasing the speed of iteration.

Also Notable

The integration of AI into corporate operations is extending beyond software development into broader strategic planning. Ferrari advises founders to aim high while maintaining rational decision-making and high talent density, noting that his previous startup, Evertail, failed after three years of pivoting, a failure he attributes partly to bureaucratic delays in Denmark. In the realm of model strategy, Jain notes that pure reasoning tasks will eventually be absorbed by model capabilities, while external relationships and tasks with low verifiable rewards will remain dependent on external tools and guardrails. This distinction highlights the ongoing tension between vertical integration by frontier labs and the survival of independent infrastructure providers. Meanwhile, the debate over AI’s economic impact continues to evolve, with Rao’s "cognitive mercantilism" framework offering a counter-narrative to the prevailing fear of human redundancy. The day’s developments underscore a market in transition, where the value of intelligence is being redefined from a scarce production input to an abundant consumer good, with significant implications for labor, marketing, and capital allocation.