MindLap

AI coding tools like Claude Code and Codex had accelerated output velocity dramatically. Teams were shipping more code, faster. But something else was happening underneath: the faster the code moved, the more the context fell apart.

Requirements lived in Jira. Architectural decisions lived in Slack threads nobody could find. Specs were written once and never touched again. The AI agents accelerating execution had no shared understanding of the system they were modifying.

The existing internal product addressed pieces of this but in isolation. MindLab was designed to be something different: not a tool that responds to prompts, but a system that actively binds intent, execution, and outcomes over time.

MindLap was the project where I understood the difference between designing for AI and designing around it. Most AI product design puts the agent at the centre and asks users to adapt. MindLap inverted that the agents are execution infrastructure, the humans are the authors, and the design job was to make that relationship legible, trustworthy, and sustainable across the life of a project. The research phase was where the real design thinking happened. Before we drew a single screen, the workshops forced us to confront the actual structural problem: not that teams lacked AI tools, but that they lacked a system that could hold the context those tools needed to operate well. Everything we built flowed from that one insight.