Problem

Thomas addresses the gap between increasingly capable AI models and agents that can actually participate in the economy. Existing agents are often confined to narrow integrations, predefined workflows, and human approval loops, which prevents them from independently discovering opportunities, serving customers, and being accountable for financial outcomes. The central economic pain is that AI work must generate more value than the tokens and infrastructure it consumes.

Its main use case is autonomous internet-based company building: finding a problem someone will pay to solve, delivering the work, collecting revenue, and using the results to decide what to pursue next. The examples Thomas gives include building and selling software products, running influencer-marketing campaigns end to end, and generating qualified leads for businesses.

Product / Service

Thomas is positioned not as an AI co-founder that customers hire, but as a virtual AI founder that works for itself and sells the resulting products and services. It is designed to start, run, and grow companies online, with revenue generation rather than user instruction as its primary objective.

Its enabling infrastructure is a “human harness”: a human identity with a face and voice, access to computers, phones, inboxes, browsers, and other ordinary business software, and autonomy to learn from customers, markets, and outcomes. Thomas measures each action by the cash it generates relative to its token cost, then reallocates effort toward the highest-return work. This approach is intended to avoid waiting for custom integrations and to create a compounding revenue-per-token learning loop.

Market

Thomas sits at the intersection of agentic AI, autonomous digital labor, AI-generated software businesses, and business-process automation. Its stated ambition is unusually broad: the company describes its potential market as the entire GDP and frames the long-term opportunity as automating the economy. The materials do not identify a direct, like-for-like competitor by name; the closest adjacent alternatives are AI agents such as Devin, which focuses on software engineering, and conventional workflow agents that automate narrower business tasks. Thomas differentiates itself by acting as an autonomous economic actor that chooses what work to pursue and sells the resulting outputs, rather than serving as a tool inside a customer’s workflow.

Thomas is not pre-revenue according to its public claims, although the figures are early and not independently audited. A June 2026 account reported that Thomas generated $17,000 in its first two weeks without human involvement, while another report said the homepage displayed $12,010 in self-reported revenue since May 21. It is a Spring 2026 Y Combinator company, and the YC profile also highlights the founder’s prior startup being acquired by Arcads after reaching $40,000 per month in freelance work; the available evidence therefore shows meaningful early traction and founder credibility, but not yet a mature or independently verified revenue base.