Back in December 2024, I had just launched Conncord and was looking for ways to grow it; basically identify new potential users so I could tell them about it.
Conncord helps businesses supercharge community driven growth by syncing their HubSpot CRM with their Discord community, enabling their sales, marketing and CSM teams to get impactful insights by de-anonymizing their community members, run powerful automations for segmentation and marketing, identify potential customers and manage customer support natively on Discord (while syncing metrics to HubSpot).
It is a fairly niche tool that servers a very small market. Because of this, identifying potential customers for Conncord turned out to be difficult. So I built a tool that would alert me whenever someone mentioned "discord" and "hubspot" in a post or a comment on Reddit. This worked very well.
Soon after that, I realized this tool had a lot of potential application and there was a serious gap in the market. And so, SnitchFeed was born.
I started building this tool mid-January of this year, and SnitchFeed has seen incremental upgrades on an almost daily basis.
At the time of this writing, SnitchFeed repos have seen a total of 817 commits. That's a whopping average of 3.19 commits/day since the initial commit on December 18, 2024.
The Most Recent Updates
Because this is the first documentation of SnitchFeed's build story, there is too much to talk about how the product has changed over the last 8 months.
I can, however, talk about what's new this month. Unfortunately, for the non-technical folks reading this, most of the updates this month have been improvements in the tech that powers SnitchFeed.
I had to focus on this rather than adding features or improving existing feature set because there was a lot of technical debt that had accumulated over the months. These technical updates to the product will make adding features and improving the product much much easier moving forward.
Switch to a mono repo
This is a technical detail—it doesn't really affect the UX or features of the tool. But it does make adding new features and polishing existing features a whole lot easier from an engineering standpoint.
SnitchFeed uses TurboRepo + pnpm for its monorepo.
Abandoning cron-based jobs in favor of more reliable and scalable worker queues
At the time of this writing, SnitchFeed is processing ~30,000 mentions every single day for its users. This is after sifting through over 100 million Reddit, Twitter, LinkedIn and Bluesky posts.
For the MVP, SnitchFeed relied heavily on cron jobs that ran every 15-30 seconds to do a lot of the processing. With growing volume of mentions (and complexity of the app), identifying and fixing bugs became a huge pain in the ass.
So I spent about a week switching everything over to worker queues. SnitchFeed uses BullMQ + Redis for this.
Right out the gate, processes are a lot more stable now, RAM usage is much lower and the app is overall more reliable. As an added benefit, we now have much better visibility into what, when and how errors occur in the app, which makes implementing fixes much less painful.
Database optimizations: Reorganizing the schema, indexing critical columns
To implement queues instead of relying on cron jobs for SnitchFeed's processes, I had to reorganize a lot of how the database was structured. New columns were created, others were nuked, and many more are now being indexed for faster data retrieval. Indexes were optimized based on the most frequent queries.
Some of this effort was more a preventative measure rather than an absolutely necessity, but I didn't want to accumulate technical debt on the database front moving forward.
SnitchFeed uses Postgres + Drizzle (❤️) for its database and ORM layer.
Final Thoughts
It has been an absolute blast building SnitchFeed and getting it to where it is today. It hasn't been all rainbows and sunshine getting here, and I never imagined it to be either.
Can't wait to discover how the product develops and grows in the coming months :)