1 minute read

Onboarding to a New Codebase with Cursor Agents and GPT-5Permalink

I recently picked up working on a new codebase, and I didn’t have much context beyond seeing what the final application should look like. I used Cursor and GPT-5 to greatly expedite my onboarding, cutting what was once a task that took many hours into something that took me 30 minutes.

Query to GPT5

Resulting doc

I then used this resulting doc as the basis for probing the repo with the help of GPT5:

Probing the codebase more Getting up-to-speed on the database layer

Next, I asked GPT-5 to learn about the coding conventions used in the repo, including the language, frameworks, static type checks enforced, testing policies, etc. I also asked it to see if there were any collaboration conventions also enforced, such as a specific ticketing system or GitHub PR rules as well as particular stakeholders to ask about the code.

Getting more context Repo coding conventions

Given all this, I had the information I needed to start working in a new codebase! What would’ve taken a couple hours a day over the course of weeks to do when starting in a new codebase, AI greatly expedites. I can now quickly get up-to-speed on a codebase, know how it works, understand what was developed and when, and have enough context to start contributing within an hour of getting started.