Subvolts take AI can now autonomously conduct research. Karpathy explains how. Mar 20, 2026 - Developers
- Learn about advanced AI agents and autonomous research systems that could shape future development tools.
- Business
- Understand how AI's increasing autonomy might impact job markets and the future of scientific discovery.
This episode features a key AI researcher discussing autonomous AI research, a significant development for the field.
Andrej Karpathy on Code Agents, AutoResearch, and the Loopy Era of AI
Audio streams from Conviction. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Andrej Karpathy joins Sarah Guo and Elad Gil on No Priors to discuss AI agents that can design experiments and improve autonomously. They explore the future of engineering and education, the impact on jobs, and Karpathy's AutoResearch project, where AI agents handle AI research loops from experimentation to optimization.
Imagine AI that can design its own science experiments and learn from them without human help. This episode discusses how AI agents are getting smarter, potentially changing how we do research and learn new skills.
This episode is for AI engineers, researchers, and anyone interested in the future of AI development and its societal impact. Karpathy, a prominent AI researcher, shares insights into autonomous AI agents and his AutoResearch project. Listeners will learn about the evolving capabilities of AI, the skills needed in the AI era, and the potential for AI to accelerate scientific discovery. This discussion is highly relevant as frontier labs like OpenAI, Google Gemini, and Anthropic are heavily investing in agentic AI and autonomous research systems.
- HostSarah Guo
- HostElad Gil
- GuestAndrej Karpathy
- Guest: Andrej Karpathy
- Show: No Priors
- Length: 67 minutes
- Project Mentioned: AutoResearch
- Discussion Topics: AI agents designing experiments, future of engineering and education, impact on jobs
What happens when AI agents can design experiments, collect data, and improve — without a human in the loop? Andrej Karpathy joins Sarah Guo on the state of models, the future of engineering and education, thinking about impact on jobs, and his project AutoResearch: where agents close the loop on a piece of AI research (experimentation, training, and optimization, autonomously).
00:00 Andrej Karpathy Introduction
02:55 What Capability Limits Remain?
06:15 What Mastery of Coding Agents Looks Like
11:16 Second Order Effects of Natural Language Coding
15:51 Why AutoResearch
22:45 Relevant Skills in the AI Era
28:25 Model Speciation
32:30 Building More Collaboration Surfaces for Humans and AI
37:28 Analysis of Jobs Market Data
48:25 Open vs. Closed Source Models
53:51 Autonomous Robotics
1:00:59 MicroGPT and Agentic Education
1:05:40 Conclusion
- Code Agents
- Andrej Karpathy
- AI programs that can write and manage computer code.
- AutoResearch
- Andrej Karpathy
- A project where AI autonomously conducts research steps like experiments and training.
- Loopy Era
- Andrej Karpathy
- A phase in AI development where systems can repeatedly improve themselves.
- Model Speciation
- Andrej Karpathy
- The process of AI models becoming specialized for different tasks or domains.
39 min1 h 1136 minSame guests, other shows
1 h 21Google Changes Search for the First Time in 25 Years
WithAndrej Karpathy
30 minAndrej Karpathy: From Vibe Coding to Agentic Engineering
WithAndrej Karpathy
2 h 25Andrej Karpathy — AGI is still a decade away
WithAndrej Karpathy
Questions people ask
- What is Andrej Karpathy's AutoResearch project?
- AutoResearch is a project where AI agents autonomously close the loop on AI research, handling experimentation, training, and optimization.
- What are the future implications of AI agents designing experiments?
- AI agents that can design experiments, collect data, and improve without human intervention could significantly change engineering, education, and the job market.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Conviction.