How Faire replaced an 18-month team project with one dev and Cursor AI

The retail marketplace Faire was staring down a massive, grueling migration of their UI stack from MobX to native React state management. A task of this scale typically sucks the soul out of an entire team for a year and a half, leaving nothing but empty coffee cups and regret.

Instead of sacrificing their human developers to the refactoring gods, the company built a custom orchestrator called Swarm on top of the AI editor Cursor. This script automatically crawls the codebase, dumps all MobX references into an Amazon S3 bucket, and unleashes parallel cloud agents in isolated virtual machines to rewrite the code.

The entire 18-month migration team was suddenly downsized to exactly one human coordinator managing a self-contained digital assembly line. While this single engineer drank tea, the automated fleet opened, tested, and merged pull requests back-to-back, leaving the rest of the company to double their overall weekly output.

Before this breakthrough, the company wasted precious engineering hours trying to build their own internal AI agent system named Samurai. Lead engineer Luke Bjerring quickly realized that hosting servers and hiring people just to maintain a fragile home-grown AI playground is an incredibly expensive way to achieve very little.

Now, the company runs more than 25 active Cursor automations, triggering over 2000 automated runs every single week without human prompts. These silent digital helpers monitor Slack for bug reports, immediately jump into the code to fix the issue, build a pull request, and even automatically repair failing continuous integration pipelines.

In another instance of speedrunning corporate bureaucracy, senior engineer Blair McAlpine wanted to build isolated preview environments for code reviews, a task he budgeted for weeks. Using Cursor's planning mode, he mapped out five interconnected pull requests, handed them to a cloud agent, and watched the AI ship the entire feature in just two hours.

Faire openly admits that buying external AI tools is vastly superior to building them from scratch, and they are now scrambling to apply this automated magic to other departments. The engineering department's productivity has surged by two to three times, making adjacent administrative and review processes the new bottleneck.

The traditional software engineering department is rapidly transforming from a room full of coders into a single supervisor managing a glowing hive of digital assistants. While junior developers sweat over their future career prospects, the tech industry is realizing that the hardest part of building software is no longer writing the code, but simply keeping up with the speed of the machines.

Source: Cursor Blog