Key Takeaways
- • millionways, a behavioral intelligence frontier lab, introduced Thorsten-4, a foundational Large Psychology Model designed to predict how humans think, decide and act, according to a Business Wire release.
- • The company reports 90% predictive accuracy for Thorsten-4 against established psychological assessments and participants' actual behavior over several years, though these figures are company-reported.
- • A multi-strategy hedge fund began trading a dedicated $50M portfolio on Thorsten-4 reads of earnings calls in Q2 2026; the company says the portfolio returned 37% in 2026 through mid-September before trading costs.
- • millionways raised $6 million in U.S. financing from a group of tech operators, investors and professional athletes, and is in use with 10+ companies across executive search, customer experience, financial services and enterprise agentic platforms.
- • Co-founder and CEO Martin Cordsmeier built the model on roughly 7,000 structured interviews conducted by 50+ psychologists through a German nonprofit foundation over a decade-long research program.
Summary
millionways, a New York-headquartered behavioral intelligence frontier lab, has introduced Thorsten-4, which the company describes as a foundational Large Psychology Model that predicts how people think, decide and act. According to the company's Business Wire release, the model is built on more than 15 years of proprietary human research and is positioned as a complement to large language models, which the company argues learn from public text and do not model why people behave as they do. The release cites a claim that ten leading LLMs fall 45% short of humans in tracking how a person's thoughts and feelings change over time, and flip their answers 46% of the time when a user pushes back. These figures are presented by the company and were not independently verified in the source material.
The model's data foundation rests on a research program led by co-founder and CEO Martin Cordsmeier through a nonprofit foundation he built in Germany. According to the release, 50+ trained psychologists conducted roughly 7,000 structured interviews with participants ranging from an eight-year-old to billionaires, and psychologists hand-tagged more than 35 million words using published psychological research methods. The company says Thorsten-4 reached 90% predictive accuracy against established psychological assessments and against participants' actual behavior over several years, without relying on self-reported data alone. The company says each output is explainable and consistent under user pushback. Companies can access the model through the millionways Platform, an API, MCP, or a self-hosted deployment.
The release highlights several applications. A multi-strategy hedge fund tested Thorsten-4 against sentiment tools and frontier LLMs and in Q2 2026 began trading a dedicated $50M portfolio based on the model's reads of earnings calls, which the company says returned 37% in 2026 through mid-September before trading costs and held up after accounting for standard market risk factors. In enterprise pilots, GPT paired with Thorsten-4 outperformed GPT alone at reading customer intent, according to the company. Multiple top-10 executive search firms confirmed the model's reads on leadership interviews against their own assessments, and millionways is developing a firm-specific assessment tool with one of them. The company is also building an Interview Integrity & Authenticity Risk API with an AI recruiting company to flag AI-assisted and coached answers.
millionways has raised $6 million in U.S. financing. Investors and backers named in the release include Ty Ahmad-Taylor, former VP of Product Marketing at Meta; Dan Gardner, co-founder and executive chairman of Code and Theory; Super Bowl champion Jason Kelce; and Pittsburgh Steelers wide receiver DK Metcalf. Scott Scanlon, CEO and co-founder of Hunt Scanlon Ventures, said his firm backed the company because behavioral intelligence will become a core AI capability. Matt Comyns, co-founder and president of Artico Search, described the technology as a major advance for executive search. Co-founder and CTO Max Weidemann framed the model as a category analogous to LLMs. The release does not disclose valuation or independent third-party validation of the hedge fund performance results.
The model's data foundation rests on a research program led by co-founder and CEO Martin Cordsmeier through a nonprofit foundation he built in Germany. According to the release, 50+ trained psychologists conducted roughly 7,000 structured interviews with participants ranging from an eight-year-old to billionaires, and psychologists hand-tagged more than 35 million words using published psychological research methods. The company says Thorsten-4 reached 90% predictive accuracy against established psychological assessments and against participants' actual behavior over several years, without relying on self-reported data alone. The company says each output is explainable and consistent under user pushback. Companies can access the model through the millionways Platform, an API, MCP, or a self-hosted deployment.
The release highlights several applications. A multi-strategy hedge fund tested Thorsten-4 against sentiment tools and frontier LLMs and in Q2 2026 began trading a dedicated $50M portfolio based on the model's reads of earnings calls, which the company says returned 37% in 2026 through mid-September before trading costs and held up after accounting for standard market risk factors. In enterprise pilots, GPT paired with Thorsten-4 outperformed GPT alone at reading customer intent, according to the company. Multiple top-10 executive search firms confirmed the model's reads on leadership interviews against their own assessments, and millionways is developing a firm-specific assessment tool with one of them. The company is also building an Interview Integrity & Authenticity Risk API with an AI recruiting company to flag AI-assisted and coached answers.
millionways has raised $6 million in U.S. financing. Investors and backers named in the release include Ty Ahmad-Taylor, former VP of Product Marketing at Meta; Dan Gardner, co-founder and executive chairman of Code and Theory; Super Bowl champion Jason Kelce; and Pittsburgh Steelers wide receiver DK Metcalf. Scott Scanlon, CEO and co-founder of Hunt Scanlon Ventures, said his firm backed the company because behavioral intelligence will become a core AI capability. Matt Comyns, co-founder and president of Artico Search, described the technology as a major advance for executive search. Co-founder and CTO Max Weidemann framed the model as a category analogous to LLMs. The release does not disclose valuation or independent third-party validation of the hedge fund performance results.
Go deeper than the headline
Research every firm in this story, track the ones that matter and get the daily briefing — free for 7 days.