Elisa Schreiber’s Post

I sat down with Peter Deng who spent years leading product at some of tech's most iconic companies—Facebook, Uber, Airtable and OpenAI. At OpenAI, he wrote the first spec for ChatGPT Enterprise and helped launch it in just seven weeks. He was instrumental in developing voice mode and worked closely with post-training teams to shape products. Now at Felicis, he's investing in the next generation of AI-native companies. In this wide-ranging conversation, Peter shares insights on: ▪️ New pricing models for AI products ▪️ How evals are becoming the new product specs in AI development ▪️ The "AI under your fingernails" test every founding team must pass ▪️ The future of voice AI ▪️ Where foundation models have advantages and where startups do ▪️ The future of product management in the age of AI ▪️ How consumer AI will fragment across multiple winners, not consolidate ▪️ OpenAI's unique culture Listen here: YouTube: https://lnkd.in/ga-tRzEe Spotify: https://bit.ly/41rm2RP Apple: https://bit.ly/41TLn78 Key Takeaways: → Why traditional pricing models are broken for AI Peter argues that seat-based pricing fundamentally misaligns with AI value creation because "the unit of work is not just the fact you get to use a tool—it's the work provided and work done." He predicts successful AI companies will move toward hybrid models combining platform fees with value-based usage, working backwards from demonstrated impact rather than forward from software access. → How evals become the new product specification At AI companies, product development transforms around a new four-person team: product, design, engineering, and post-training leads working as one unit. Peter explains that "the new spec is the eval" because defining AI behavior requires measurable evaluations rather than traditional feature descriptions. This fundamental shift means product managers must learn to specify model behavior through testable criteria, making evals the critical bridge between product vision and model capabilities. → The "AI under your fingernails" founder advantage Peter uses the analogy of humanity discovering "a new element in the periodic table" to describe AI capabilities. The founders who succeed are those who've experimented deeply with models and understand their raw properties firsthand. This hands-on experience becomes more valuable than pure domain expertise, as it enables founders to discover capabilities that even model creators haven't fully explored. → Voice AI's unique cognitive unlock Beyond hands-free convenience, voice enables a fundamentally different interaction mode with AI. "When you form words, the brain is forming thoughts at the same time," Peter explains, removing the barrier between thought formation and expression that exists with text interfaces. This creates opportunities for real-time collaborative thinking.