Andrej Karpathy Joins Anthropic: What Happens Next

2026-05-19 · Source: The Algorithmic Bridge · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Robotics & Autonomous Systems, Emerging Technologies & Innovation · Depth: Advanced, extended

Summary

Andrej Karpathy, a founding member of OpenAI and former Tesla Autopilot lead, joined Anthropic on May 19th, 2026, to work on pre-training, specifically using Claude to accelerate recursive self-improvement (RSI) research. This move follows his earlier skepticism about AI agents, which shifted dramatically after the "agentic boom" sparked by Anthropic's Claude Code in late 2025. Karpathy's independent "autoresearch" work, focused on automating AI research using AI agents, aligns with Anthropic's goal of achieving autonomous AI R&D, with co-founder Jack Clark predicting a 60% chance of full RSI by the end of 2028. His decision underscores a growing industry focus on AI systems improving themselves.

Key takeaway

For AI leaders and researchers, Karpathy's decision to join Anthropic highlights a critical industry focus on recursive self-improvement and AI-driven research. This shift, driven by advancements like Claude Code, suggests that future AI development will increasingly rely on autonomous agents. You should prioritize integrating AI agents into your research and development workflows to stay competitive, focusing on structured tasks and leveraging AI for efficiency gains in model pre-training and optimization.

Key insights

Karpathy's move to Anthropic signals a critical industry shift towards AI-driven recursive self-improvement and automated research.

Principles

• AI progress involves a "march of nines," where each improvement level requires constant effort.
• Education should build "ramps to knowledge" for efficient, motivated learning.
• Human intelligence relies on generalizable patterns, not just memorization.

Method

Automating AI research involves using AI agents to collaboratively tune models, promoting promising ideas to larger scales, with optional human contribution on the edges.

In practice

• Use autocomplete for coding, reserving LLM agents for boilerplate or unfamiliar languages.
• Prioritize learning depth-wise and on-demand for project-driven rewards.
• Explain concepts to others to deepen personal understanding and identify gaps.

Topics

• Andrej Karpathy
• Anthropic
• Recursive Self-Improvement
• AI Agents
• LLM Pre-training
• AI Research Automation

Code references

• karpathy/autoresearch

Best for: Research Scientist, AI Scientist, Director of AI/ML, Investor

Related on AIssential

• Prominent AI researcher Andrej Karpathy picks Anthropic over former home OpenAI to get back into frontier LLM research — Andrej Karpathy's move to Anthropic signals a strategic bet on AI self-improvement for foundational LLM pretraining.
• Anthropic Says AI Is Already Building AI (And the Numbers Back It Up) — Anthropic's Claude AI now authors over 80% of its own codebase, signaling a present-day inflection point in recursive self-improvement.
• This is absolutely CRAZY — Andrej Karpathy joining Anthropic validates a pessimistic AI future, highlighting the company's safety-first, closed-source philosophy.
• Anthropic just disclosed that Claude now writes more than 80% of the production code it merges — AI agents are dramatically increasing developer productivity and code generation capabilities, while raising recursive self-improvement concerns.
• Why AI Is Replacing Code Reviews Faster Than Anyone Expected — AI models like Opus 4.5 are rapidly shifting from collaborative tools to autonomous code generation agents.
• OpenAI VS Anthropic — Frontier AI labs are rapidly advancing generative coding, focusing on agentic systems and self-improvement.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Algorithmic Bridge.