Why Merge CEO Shensi Ding stays hands-on with Codex

**TL;DR:** Merge CEO Shensi Ding writes code herself while managing the company’s non-engineering operations. She shared her astonishing first experience with coding AI tools (Codex / Astra) and how they enable colleagues without engineering backgrounds to build their own workflows. ## One Core Idea: Executives Set the Ceiling for AI Proficiency "If you’re an executive, you are the ceiling for AI proficiency across your entire organization." This was the starting point for Shensi Ding’s reflections. As Merge’s CEO, she manages the company’s non-engineering operations, but she also writes a fair amount of code. She does not believe leaders can ask their teams to embrace AI while staying on the sidelines themselves—executives’ familiarity with these tools directly determines how far the organization can go in applying AI. ## Her First Experience Using a Coding AI Tool Shensi Ding revealed that before Asper was released, she had never used a coding AI tool, and her first experience truly amazed her. At the time, she was working on redesigning a feature but was unsure which approach would work best. So she asked Astra directly: > "Can you draw some inspiration from other parts of the product, get a sense of what our design system looks like, and then give me your best proposal?" The team ultimately built a feature based on Astra’s proposal. ## The Key Difference: Remembering Every Requirement, Rather Than Just Following the Last One In her view, Astra’s biggest difference from other AI systems is its ability to handle context. She said that other AI systems sometimes forget earlier requirements and only carry out the last thing she mentioned. Astra, by contrast, works through every task she gives it and executes each one—and does so exceptionally well, using sound reasoning to fill in details she has not explicitly stated. This ability to "infer unspoken requirements" is what she believes makes the tool truly useful—users do not have to turn every step into an airtight instruction. The AI fills in the gaps based on its understanding of the product and design system. ## What This Means for Colleagues Without Engineering Backgrounds Shensi Ding believes this is especially helpful for team members without engineering backgrounds, because they can now build almost anything. They no longer have to wait for colleagues to help them. As soon as they think of a workflow that would be particularly useful to them, they can start building it themselves. She said this is truly wonderful. This also echoes her opening point: only when managers and people in non-engineering roles can use these tools themselves does AI proficiency truly become an organizational capability, rather than something exclusive to the engineering team. Source: Why Merge CEO Shensi Ding stays hands-on with Codex — OpenAI (https://www.youtube.com/watch?v=6ep8Vvm4Bdw)