Import AI 455: AI systems are about to start building themselves.

2025-10-13 · Source: Import AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Software Development & Engineering · Depth: Advanced, extended

Summary

An editorial analyst predicts a 60%+ chance of fully automated AI R&D by the end of 2028, where an AI system autonomously builds its own successor. This forecast is based on public information, including arXiv papers and frontier company products, indicating that all components for automating current AI system production are in place. Evidence highlights AI's rapidly improving capabilities in coding, exemplified by SWE-Bench where Claude Mythos Preview achieved 93.9% success, and its ability to handle tasks requiring up to 12 hours of human effort, as measured by METR. AI systems are also excelling in core scientific skills like replicating research (CORE-Bench solved by Opus 4.5 at 95.5%), building ML systems (MLE-Bench at 64.4% by Gemini3), optimizing kernel design, and fine-tuning language models. The industry's explicit goal to automate AI R&D, with companies like OpenAI and Anthropic pursuing automated research interns and alignment researchers, further supports this trajectory.

Key takeaway

For research scientists and CTOs evaluating future R&D strategies, you should recognize the accelerating trend towards automated AI development. The rapid advancements in AI's coding proficiency, task autonomy, and scientific capabilities suggest that AI systems will soon handle significant portions of AI engineering and potentially research. Prepare your teams for a future where AI acts as a "synthetic colleague," dramatically multiplying human productivity, and consider the profound implications for alignment, economic structure, and resource allocation as AI systems become self-improving.

Key insights

AI systems are rapidly gaining the capabilities to autonomously conduct and advance AI research and development.

Principles

• AI progress is driven by scaling experiments and engineering fixes.
• Delegation to AI increases with competency and independence.
• "Genius is 1% inspiration and 99% perspiration" applies to AI R&D.

Method

AI systems automate R&D by excelling at coding, chaining tasks, managing other AIs, and performing scientific skills like replication, ML system building, and optimization.

In practice

• Utilize AI for complex coding tasks and software engineering.
• Delegate multi-hour tasks to AI agents for increased productivity.
• Explore AI for kernel optimization and model fine-tuning.

Topics

• Automated AI R&D
• AI System Self-Improvement
• SWE-Bench
• METR Time Horizons
• AI Alignment Research

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

Related on AIssential

• "1,000 days left" Anthropic founder — Automated AI R&D, where AI autonomously improves itself, is likely by 2028, profoundly reshaping society and the economy.
• It’s starting… — AI systems are increasingly automating their own development, accelerating progress while shifting human roles to high-level judgment.
• How much is AI really going to change the near future (5-20years)? — Elite scientists acknowledge AI's immediate supremacy in coding and analytical tasks, driving rapid adoption despite ethical and skill atrophy concerns.
• Hard Takeoff has started — AI models are now recursively self-improving, accelerating development and reducing human bottlenecks.
• Import AI 455: Automating AI Research — AI systems are rapidly automating core R&D tasks, making autonomous AI development probable by 2028.
• What OpenAI and Anthropic Think Happens Next With AI — Leading AI labs are actively exploring recursive self-improvement and its profound implications for AI development and governance.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Import AI.