Better AI Models Won't Fix Your Engineering Team — Nicholas Arcolano, Jellyfish

Nicholas Arcolano, head of AI and research at Jellyfish, uses data from about 50,000 engineers at 447 companies to ask how much the model really matters for AI coding output. When Cursor pushed many of its users onto Grok, merged PRs barely moved. Breaking down what drives throughput, the tool explains under 1%, skill use about 3% (heavy skill users see roughly a 30% bump), token usage about 9%, company and team about 12%, and the individual engineer most of the rest. The engineers who were top performers before AI are mostly still the top performers now. His advice for leaders: focus on the people, skills and environment you can control. Chapters: - 00:00 Intro - 00:40 A golf analogy - 02:01 Models keep getting better - 02:59 A natural experiment: the Grok switch - 04:54 What actually drives output - 07:14 The tool - 08:07 How well you use it: skills - 10:00 How much you use it: tokens - 11:23 Where you work - 12:22 You, the individual - 13:42 Is AI reshuffling the top engineers? - 14:50 What leaders can control Nicholas Arcolano, Jellyfish: https://jellyfish.co Join us in SF for the second annual AIE CODE summit: https://aie.codes our last flagship summit of 2026. Applications live now!