Import AI 448: AI R&D; Bytedance's CUDA-writing agent; on-device satellite AI
A credible academic framework for measuring AI R&D automation offers APS policy teams a structured basis for thinking about AI oversight reporting obligations.
Key points
- GovAI and Oxford propose 14 measurable metrics to detect progress toward AI recursive self-improvement.
- The framework explicitly calls for government access to confidential industry reporting on AI R&D automation.
- Remaining items cover ByteDance's CUDA-writing agent, edge AI for satellites, and an AI timeline update - context only for APS readers.
Implications for Australian agencies
- Monitor Policy teams tracking AI governance frameworks may want to monitor the GovAI/Oxford AIRDA metrics paper as a potential input to future mandatory reporting or procurement risk frameworks.
- Consider Agencies developing AI risk or frontier AI governance positions could consider whether the 14 AIRDA metrics offer a structured lens for assessing vendor AI development practices.
Implications are AI-generated. Starting points, not advice — see methodology for how they're framed.
View original source
Copied.
Appeared in:
Weekly digest, 9 March 2026
"Import AI 448: AI R&D; Bytedance's CUDA-writing agent; on-device satellite AI"
Source: Import AI – Substack (Jack Clark)
Published: 9 March 2026
URL: https://importai.substack.com/p/import-ai-448-ai-r-and-d-bytedances
This edition of Import AI covers four technical research items. The most governance-relevant is a GovAI and University of Oxford paper proposing 14 metrics to measure AI R&D Automation (AIRDA) - the degree to which AI systems are building successor AI systems - and recommending that governments develop confidential industry reporting mechanisms to track this. A separate item covers Ajeya Cotra's updated AI capability timelines, which now forecast agent task horizons exceeding 100 hours by end of 2026. ByteDance's CUDA Agent demonstrates AI systems writing GPU kernel code, and German researchers present TinyIceNet, a miniaturised vision model for satellite-based sea ice monitoring. A speculative short story on autonomous drone warfare closes the issue.
Implications for Australian agencies:
- [Monitor] Policy teams tracking AI governance frameworks may want to monitor the GovAI/Oxford AIRDA metrics paper as a potential input to future mandatory reporting or procurement risk frameworks.
- [Consider] Agencies developing AI risk or frontier AI governance positions could consider whether the 14 AIRDA metrics offer a structured lens for assessing vendor AI development practices.
Retrieved from SIMS, 18 July 2026.