Google DeepMind Says Self-Improving AI Remains a Distant Prospect
Understand the story
On October 15, 2026, The Decoder reported that Google DeepMind engineer Haydn Belfield said AI-driven acceleration of AI research has not yet emerged, and self-improving AI remains far off. Internal data he provided shows: AI now writes more than 75% of production code, but about 88% of code changes still require human oversight or review; the share of code changes fully delegated to AI rose from 7.5% to 11.9%. Meanwhile, the more complex the task, the lower the AI's independent completion rate. His remarks indicate that although AI's share of code writing is already high, it still has clear limitations in autonomy and independently completing complex tasks, and achieving self-improving AI remains a long-term goal.
AI overview of reporting · Updated 4 days ago
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderGoogle says self-improving AI isn't anywhere close
Google DeepMind engineer Haydn Belfield said AI-driven acceleration of AI research has not yet appeared and self-improving AI remains far off. AI now writes over 75% of production code, but roughly 88% of code changes still require human oversight or review; the share of code changes fully handed to AI rose from 7.5% to 11.9%, and the more complex the task, the lower the rate of independent AI completion.
Story heat trend
Current heat3·Comparable peak10（10/15 04:00）·Comparable change over 24 hours -50%
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.