At Amazon, there is a famous tradition designed to keep builders grounded: always keep an empty chair in the meeting room. That chair represents the customer, serving as a silent, constant reminder of the real people who will eventually use whatever software, feature, or product is being designed. Today, as generative artificial intelligence accelerates software development to unprecedented speeds, keeping that human perspective at the center of the process has become incredibly difficult. Three Amazon veterans—Rohit Talluri, Jean Farmer, and Gabriel Fong—recognized that in the rush to ship AI-generated code, the actual human consumer is easily forgotten. This realization inspired them to launch Primitive Labs, a San Francisco-based startup aiming to pull humanity back into the loop by giving AI agents a permanent seat at the development table.
Primitive Labs is pioneering a concept they call “behavioral intelligence.” The startup builds highly specialized AI agents designed to observe, reason, and act precisely as human customers would across various software platforms and digital devices. Instead of relying on traditional, slow-moving user research, focus groups, or surveys—which can take weeks or months and are often skipped by fast-moving product teams—Primitive Labs automates the feedback loop. By running new designs, features, or marketing choices through these lifelike digital simulators first, product teams can instantly see how real people will react before writing a single line of public code. The ultimate mission, as the founders put it, is to turn human behavior into a “first-class primitive” of the software development lifecycle, ensuring that modern technology remains deeply intuitive, trustworthy, and aligned with human needs.
The talent driving Primitive Labs brings deep, enterprise-grade AI experience to the table. CEO Rohit Talluri previously worked in Amazon’s prestigious AGI Autonomy Lab, where he built computer-use agents alongside industry pioneers and helped launch Nova Act, Amazon’s flagship agentic model. Chief Technology Officer Jean Farmer also transitioned from Amazon’s AGI group, where she specialized in tool-use capabilities for the Amazon Nova models, designing the frameworks that allow AI to interact seamlessly with external software. Chief Operating Officer Gabriel Fong, who has partnered with Talluri since 2020, brings a strong background from Amazon Web Services (AWS) and DoiT International, managing everything from product direction and customer discovery to hands-on technical operations. Together, this founding trio is leveraging their background in massive machine-learning infrastructure to build agents with unmatched “behavioral fidelity”—meaning the virtual agents’ choices closely mirror real human decisions.
This unique vision has quickly captured the attention of top-tier Silicon Valley investors. Primitive Labs recently closed an undisclosed pre-seed funding round led by a16z Speedrun, the highly competitive accelerator program run by Andreessen Horowitz. They are supported by a stellar lineup of venture funds and angel investors, including Olive Tree Capital, Cloverfield Fund, and Unexpected Investments. High-profile individual backers include former TechCrunch editor Josh Constine, former Adept CEO David Luan, and several AI experts with deep roots at OpenAI, Google DeepMind, Databricks, Nvidia, and Meta. The startup is currently operating out of Andreessen Horowitz’s Speedrun space in San Francisco, with plans to establish their own permanent headquarters in the city as they hire their founding engineers and researchers. For Talluri, a University of Washington graduate who once dreamed of launching a startup while reading local tech news, setting up shop in San Francisco was a deliberate choice driven by the city’s dense concentration of AI talent and rapid research pace.
On the product front, Primitive Labs is already running private previews with a select group of early clients, including household names from the Fortune 50 and Fortune 500 lists in the consumer technology and e-commerce sectors. While their current focus is refining digital customer journeys across mobile phones and computers, the team has even explored using their cognitive models to predict reactions to physical designs, such as product packaging and branding. The underlying technology blends advanced computational cognitive science with custom, human-like memory systems that allow agents to retain information and learn over time. Committed to advancing the broader AI ecosystem, the founders plan to publish their scientific findings and open-source portions of their core behavioral architecture over the coming months as they march toward a general public release later this year.
Ultimately, Primitive Labs wants to redefine how software is built in the age of automation. Rather than simply using AI to write code faster, they want to use AI to make technology more empathetic and user-friendly. When asked if the team plans to honor the old Amazon tradition of keeping an empty chair in their new office, Talluri did not hesitate to say yes, but with a modern twist. They will absolutely keep a physical chair open in their conference room, and when they look at it, they will be envisioning one of their intelligent behavioral agents sitting there, representing the diverse voices of millions of customers worldwide.
