John Schulman Discusses Recursive Self-Improvement at Thinking Machines
The interview signals that Thinking Machines is positioning itself around AI-accelerated research rather than any single product, a useful signal for reading the lab's roadmap. Researchers and lab watchers who track frontier-lab strategy will get Schulman's first detailed public framing of how he thinks measurement and human judgement fit into self-improvement loops. It also offers a concrete comparison point against similar autonomist statements from OpenAI and Anthropic.
Thinking Machines co-founder John Schulman sat down with Sarah Ding Wang for a conversation on RSI (recursive self-improvement): how AI can accelerate AI research itself. The discussion, published on the company's X account, centres on three core questions: how to measure AI's contribution to research acceleration, the concrete mechanisms through which that acceleration operates, and what role human taste and intuition play in the loop.
Schulman co-founded OpenAI and served as its VP before leaving in 2024 to join Thinking Machines. Framing the conversation around recursive self-improvement serves as the company's public position on AI-driven autonomous research, a direction it wants associated with its brand.
Recursive self-improvement has become a recurring theme among frontier labs, and public discussions like this one signal how companies justify research autonomy claims to both the technical community and the broader public.
Our own @johnschulman2 sat down with @sarahdingwang to talk all things RSI: how AIs accelerate progress, how we measure it, and where human taste and intuition fit in the picture.
This report is categorized under AITOP's "Industry" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include Thinking Machines, John Schulman, RSI, AI safety.
Source attribution: curated from Thinking Machines Lab (@thinkymachines).