How AI-assisted development helped us build the new Dropbox experience
Pravi Garg, Sourabh Yerfule and Gennadiy Civil · 5 November 2026
Read the originalDropbox engineers handed coding agents defined tasks, each starting from a written plan and ending in an engineer's review, to finish a product launch on time.
Pravi Garg, Sourabh Yerfule and Gennadiy Civil wrote on the engineering blog of Dropbox, the file storage company. They described how engineers built its new AI experience for web and desktop against a November launch deadline. Engineers moved from using AI for single coding steps to delegating defined tasks, including implementation and testing, for engineers to review.
- Each delegated task began with a written plan agreed by the engineer and the AI coding agent, and the agent ran builds and tests and recorded the results for review.
- Instructions stored with the code set the rules agents follow, and a working group with champions across teams set shared task formats and review expectations.
- A cleanup of automated tests that had been deferred for two quarters was split into 53 tasks and finished in one week.
- Agents also tried the product during development and found a deleted chat item that stayed visible until the page was refreshed.
An engineering team can delegate more work to coding agents when every task follows the same written plan, the same test run and the same review step. This is an inference and is not the authors' claim. The piece is Dropbox's account of building its own product.
Split long-deferred cleanup work into small defined tasks that coding agents complete through the same test and review process.
Start each task given to a coding agent with a written plan agreed by the engineer and the agent, and have the agent record its test results for review.
Derived by Working Surface from the article; more in the Playbook. Source line: Instructions stored with the code explained the rules agents needed to follow, giving engineers a repeatable way to delegate work while remaining responsible for the result.
6 November 2026: teams scaling AI agents are replacing case-by-case checking with one repeatable path that every piece of agent work passes through.