4-month stipend: Research on agent scaling laws—relationships between training compute and agent capabilities of LLMs | grantmaking.ai