Humans& launches Persimmon, first large-scale "user model" simulating human conversation

On September 11, AI startup Humans& (Humansand) officially released Persimmon, which it calls the first large-scale "user model," purpose-built to realistically simulate how humans converse and interact. The project is led by 314Bansal and alexisjross, and the team says it is the result of a long period of refinement.

Confirmed

• Humans& positions itself as an AI alignment company, with a core belief that for AI to coexist and collaborate with humans, it must first genuinely understand humans—including representing people as they actually are, not just as we wish to be seen.
• Team member saurabhshah2 explained that these simulated user models are "quirky, irritable, and spontaneous, just like real people," and the most interesting outputs often come from that authenticity.
• The company says user models let AI reason through hypothetical scenarios and explore how an event or decision might play out.

Why it matters

• If the simulations are realistic enough, Persimmon could provide a controllable "human behavior sandbox" for AI alignment research and product testing, reducing the cost of relying on real-user testing.
• Multiple sharers (@niloofarmire, @Scobleizer, and others) noted its "understanding humans" positioning, viewing it as a new attempt in the alignment field.

2026-09-11 ~ 2026-09-11 · 10 related posts

• [source] Humans& launches Persimmon, first large-scale model to simulate human interaction — niloofar_mire · 2026-09-11
• Noah Goodman's humansand launches Persimmon, a model built to simulate human interaction — gharik · 2026-09-11
• [source] Humansand launches Persimmon, a model built to simulate human conversation that fools AI detector Pangram — gharik · 2026-09-11
• Humans& Launches Persimmon, a User Simulation Model Built on NVIDIA Nemotron — niloofar_mire · 2026-09-11

6 near-duplicate retellings: niloofar_mire · niloofar_mire · Scobleizer · niloofar_mire · gharik · xiaosun86