Summary
In this episode of the Light Cone from Y Combinator, the hosts talk with Lauren Tan and Roman Ugarte, members of the core team behind Grok Bot, a product for creating AI teammates. The team describes taking it from first line of code to launch in about seven weeks, starting as an internal prototype built after seeing how long-running coding agents transformed their engineers’ work.
The conversation covers the design principles behind the product: work is organized around bots rather than chats or files, bots get the same tools humans use, including their own computer, identity, phone number and credit card, and they are always on and proactive rather than waiting for prompts. The team notes that both heavy AI users and the least AI-forward employees took to it internally.
They share examples of how teams use it, such as a performance marketing bot given access to Google Ads and Facebook Ads with a budget, collaborating with a design team’s bot to A/B test assets. One walkthrough shows a feature built end to end in a single 200-message Slack thread, with the bot spinning up a Cursor cloud agent to prototype and return screenshots and videos. The discussion also looks at what codebase guardrails and tooling make this style of development work, and what software engineering may look like in the coming year.
📺 Source: Y Combinator · Published November 14, 2026
🏷️ Format: Interview







