NYTech
People
Personfounder

Courtland Leer

Courtland Leer is a co-founder of Plastic Labs, a research-driven artificial intelligence company, where he serves as chief operating officer, a role that has also been described publicly as president of the company [1][2][3]. He holds a Bachelor of Arts in Philosophy from The University of the South, and has been with Plastic Labs since its founding in January 2023 [2]. Leer has used media appearances to articulate the company's mission as an attempt to solve what he frames as a version of the economic "principal agent problem," the informational and incentive asymmetries that arise between a person and any good or service provider, arguing that reducing these asymmetries is what enables better collaboration between humans and AI systems [5]. He contends that alignment in large AI systems today is imposed top-down, with major labs tuning their models to broad corporate values in order to homogenize the experience for the widest possible audience, a practice he argues produces poor user experience and excessive restriction; Plastic Labs' stated mission, in his account, is to "decentralize alignment" so that agents can instead be tuned to individual users, organizations, or communities [5].

Under Leer's leadership, Plastic Labs built Honcho, a hosted platform intended to give AI agents memory, context, and social cognition specific to each user, which he describes as requiring an "ambient" way of gathering data on an application's end user to build a synthetic, evolving representation of that person, which can then be supplied to an AI system at the moment of inference [5]. Leer has stated that achieving this kind of personalization is largely a context problem, in that an agent's ability to align with a person's interests depends on having a mechanism to learn about and store information on that individual and inject it into its own reasoning process [5]. He has also distinguished Plastic Labs' focus from foundation-model development, saying the company works on post-training alignment techniques rather than pretraining large models itself, while devoting substantial internal effort to tracking outside research literature, citing the hiring of a machine learning research engineer as the company's first full-time hire as evidence of this research orientation [5].

Leer has described a "peer paradigm" as an organizing concept behind Honcho's architecture, arguing that most existing memory and interaction systems assume a single user talking to a single assistant, a model he says cannot support more complex arrangements such as group chats or multi-agent systems [6]. In Honcho's design, he has explained, any entity capable of holding a changing context, including an individual user, an AI agent, an organization, or a group, can be treated as a "peer," and an application can synthesize context by drawing on multiple such peers at once [6]. He has pointed to areas such as agent-controlled cryptocurrency wallets, where an AI agent rather than its developer holds and autonomously transacts with funds, as an example of the more autonomous, trust-requiring uses he expects personalized alignment to eventually support [5]. Leer discussed these ideas publicly around the announcement that Plastic Labs had raised a $5.35 million seed round led by Variant, White Star Capital, and Betaworks, with participation from Mozilla Ventures, Seed Club, Greycroft, Differential Ventures, and a group of angel investors [5][7][8][9][10][11].

Founded

Insights & ideas

Courtland Leer frames Plastic Labs' mission as solving the principal agent problem, the old economic issue of information and incentive asymmetries between parties [2]. He argues that current AI alignment is imposed top down by large labs according to corporate values, producing homogenized, frustrating user experiences, and that trustworthy autonomous agents instead require alignment that is decentralized and built bottom up, tailored to individuals, organizations, or communities [2]. This underpins Honcho, described as a personal identity layer giving agents memory, context, and social cognition specific to each user, including preferences, personality, and values, so agents can model who someone actually is rather than just recall past sessions [2]. He also discusses agent-owned wallets and autonomous control over funds as part of a broader convergence of crypto and AI around data sovereignty [2]. Beyond one-to-one chat assistants, Leer advocates a peer paradigm where agents, users, groups, and organizations can all function as peers, enabling multi-agent systems and group interactions [1].

Education

Media & appearances

This page shows public professional information only, each fact cited. Is this you? Corrections or removal within 24 hours, no questions asked.