Shipping Threads in 5 Months

As developers, we often prefer writing new code rather then reusing the older one. Of course, new code will be better, faster, more readable, maintainable, it will use newer tools and frameworks and definitely will be great. Spoiler: No 😉

Old battle-tested code can be significantly better because it's already tested, covered by automations, it has common features in place, little learning encoded, it doesn't have stupid mistakes already, it's mature.

That's why I think there is really interesting experience that shared by Meta team how they reuse Instagram code to build Threads app from scratch.

Initially, the team had an ambitious goal to build a service to compete with X (Twitter) in a couple of months. To achieve that, they decided to use existing Instagram code with core features like profiles, posts, recommendations, followers as the base for the new service. Additionally, Threads was built on existing Instagram infra with support of existing product teams. This approach allows to deliver new fully-featured service in 5 months.

Key findings during the process:
✏️ You need the deep knowledge of legacy code to successfully reuse it
✏️ Code readability really matters
✏️ Repurposing and customization existing code for the new requirements brough additional technical debt that should be paid in the future
✏️ Shared infrastructure and product teams can significantly reduce development costs
✏️ Old code is already tested in real conditions and contains less issues then the new code

So don't rush to rewrite the existing code. Thoughtful evolutionary approach can bring more business benefits, reducing time to market and overall development costs.

#engineering #usecase