a story in 3 parts... on friday, the team was concerned that we were building too many integrations on top of existing ESPs (e.g. braze, iterable) and not spending enough time thinking about our own platform. on saturday, i wrote an internal blog post explaining how today's ESPs work and reverse-engineering what their architecture probably looks like based on what i could find from public info/blogs/APIs + inspecting network calls. tldr, it's a lot of kafka + elasticsearch/mongodb. on sunday, i built a simplified MVP on @convex @tan_stack and @CloudflareDev for this hackathon. it actually fucking works, which is pretty cool, and we still have until 11/17 to win up to $100K in cloudflare credits, which i desperately want. @tannerlinsley @jamwt @threepointone , as judges... 1) do you think this is cool? 😅 2) if i have a couple more weekends to jam on this, do you think i should focus on productionizing for real-world workloads or being more experimental but less practical? for context, our clients are lifecycle marketers who manage email campaigns for B2C companies, so the scale we'd need to support would be in the MMMs.
