Andrew Chen compared two product mandates: in 2012, it was to "add chat or messaging to mobile applications"; in 2026, it is to "add note-taking to AI applications." The post did not elaborate further.

Around 2012, social and tool applications treated instant messaging as the default growth module: retention relied on conversations, network effects depended on inboxes, and daily active users were driven by notifications. Chen was known for his research on growth and cold starts, later joined a16z, authored a thesis on network effects and cold starts, and participated in discussions on growth for ride-sharing platforms. By 2026, a similar list emerged for AI applications: beyond chat windows, features like audio transcription, meeting minutes, personal knowledge bases, and memory layers were added, allowing model outputs to be consolidated into retrievable notes. The a16z consumer group wrote earlier that year about knowledge work tools being biased towards execution and lacking thinking tools, with the model itself treated as a thinking partner; in March, Chen stated that the value of proxy products lies in closing a row of browser tabs, with the number of tabs serving as a proxy metric for compressible workflows.

Note-taking capabilities have appeared in memory and project spaces of chat products, generation pages of document tools, meeting assistants, and local note-taking software, with overlapping functional boundaries. The additive list does not automatically create network effects: chat in 2012 connected people to people, while notes in 2026 connect individuals to their own context. The post did not name any specific companies or provide market size.

Who is buying and who is selling: the event is driven by a cycle of functional lists, not by the launch of new protocols. Buyers are startup teams aiming to turn AI applications into "all-in-one workstations," while sellers are providing the same transcription and vector libraries to all chat shells as infrastructure. Beneficiaries are existing distribution entry applications, while tools that only focus on note-taking are under pressure. Funding is shifting from model inference fees to context storage and retrieval, while user attention is split among multiple "note-taking chat boxes."

Source: Public Information