Greg Isenberg Releases 20 Observations After Returning from San Francisco, Pointing Out the Emergence of Terms like Inference on Local Billboards
Greg Isenberg released 20 observations after returning from San Francisco, noting that local billboards have begun to feature technical terms like inference, while the vast majority of Americans still do not understand the concept of Agents. The entrepreneurial ecosystem is shifting from showcasing the number of hires to flaunting revenue per capita.
Young people under 22 are earning incredible amounts of money, and scenarios where small teams operate companies with a large number of Agents are becoming increasingly common. Founders are no longer proud of the size of their workforce but rather of generating higher income with fewer people as a new bragging metric.
In March 2025, Manus launched an interactive paradigm used by many Agent products today, assigning a cloud computer for each task, allowing humans to observe the execution process, with the system automatically generating a task list and completing it item by item, enabling users to leave their computers and wait for results.
Products now must undergo two sales: the first by the Agent selecting the product, and the second by humans trusting and confirming the Agent's choice. The customer acquisition, marketing, pricing, and payment processes designed around human users are being redefined.
AI has increased the efficiency of freelancers and agencies by at least five times, yet many still charge by the hour. Clients will eventually discover that the same work only requires one-fifth of the original time, forcing a reevaluation of pricing models. Meanwhile, employees are independently using AI to develop internal tools, and company IT departments can no longer keep track of the number of software running internally.
In market mechanisms, entrepreneurs willing to use Agents to reduce labor costs are buying efficiency, while traditional hourly-rate service providers and standardized SaaS vendors are facing pressure. Funding is flowing towards smaller teams with higher per capita output, while some members of the public oppose data centers, calling for their outright destruction. AI companies are beginning to adjust their narratives to emphasize curing cancer and increasing free time.
Only 2% of Americans are paying for AI products, but San Francisco feels like 110% of people are using them. The proactive features of Agents, such as discovering flight delays and completing rebookings, are being discussed as potential directions for 2027. Entrepreneurs believe that the current period represents the most asymmetrical window commercially.
Source: Public Information
ABAB AI Insight
Greg Isenberg has long observed the entrepreneurial ecosystem in Silicon Valley, previously recording changes in founder behavior during the mobile internet and Web3 phases. This time, he focuses on how Agents are reshaping organizational forms rather than just technical parameters, continuing the historical path from "competition in workforce size" to "competition in capital efficiency."
On the capital path, funding is shifting from large-scale hiring and standardized SaaS procurement to a combination of small teams and Agents. Founders are increasing revenue per capita by reducing headcount, while tool providers like Manus gain first-mover advantages by defining Agent interaction standards. Resources are moving from human labor time to replicable cloud execution environments, motivated by the desire to lower marginal costs and accelerate product iteration.
In analogy, this is similar to the early stage of the mobile internet era in the 2010s, where "small teams relied on app distribution," but the current pace is faster, with Agents replacing internal execution rather than just customer acquisition. The industry is transitioning from human-driven SaaS to software selecting software, and the explosion of personal software is akin to the early proliferation of Excel macros and internal tools, now directly generated by AI.
Structurally, this represents a reconstruction of the industrial chain, with Agents beginning to serve as an intermediary layer for procurement and decision-making, making the first customers of software companies other software rather than humans. The mechanism involves a shift in trust from humans towards brand loyalty to trust in the logic of Agent choices, with pricing power moving from hourly or per seat to results-based or based on Agent invocation volume. Meanwhile, social sentiment fragmentation is forcing AI companies to shift their narratives from showcasing capabilities to committing to social value.
ABAB News · Cognitive Laws
Software is first sold to software, then to humans.
Revenue per capita replaces headcount as the new bragging metric.
Efficiency increases fivefold, yet pricing is still calculated by the hour.