Loading video...
Video Failed to Load
We've been obsessing a lot about creating the best possible AI support experience Dub.co. And I think we've found the magic formula. The stack: ✦ Mintlify → docs + help articles ✦ Upstash vector → vector search ✦ AI SDK + Claude sonnet → LLM/retrieval ✦ Plain → support... show more
23,029 views • 7 months ago •via X (Twitter)
26 Comments

Try out this experience for yourself – on both and 😁 Kudos to our new hire @pedrooladeira for giving our support experience the upgrade it needed!

@dubdotco @mintlify @upstash @aisdk @claudeai lol literally doing the same thing

@dubdotco @mintlify @upstash @aisdk @claudeai LFG 🚀 great minds!

@dubdotco @mintlify @upstash @aisdk @claudeai Just tried it out, it has knowledge of my account and help docs without even speaking to a human?! Soooooo good! Crazy impressed!

@dubdotco @mintlify @upstash @aisdk @claudeai Let’s goooo! Glad you liked it 😁

@dubdotco @mintlify @upstash @aisdk @claudeai looking good 💅

@dubdotco @mintlify @upstash @aisdk @claudeai Mintlify’s Aspen theme is soooo freaking clean 🤩

@dubdotco @mintlify @upstash @aisdk @claudeai shoutout @branmcconnell

@dubdotco @mintlify @upstash @aisdk @claudeai repo 🥲?

@dubdotco @mintlify @upstash @aisdk @claudeai all OSS @ 😉

@dubdotco @mintlify @upstash @aisdk @claudeai

@dubdotco @mintlify @upstash @aisdk @claudeai gold standard 🥇

@dubdotco @mintlify @upstash @aisdk @claudeai 🥹🫶appreciate you, will keep improving it!!

@dubdotco @mintlify @upstash @aisdk @claudeai solid stack. the tool() approach for workspace data is clever - gives grounded responses without bloating context the sync-on-push to vector DB is what most teams miss. AI support breaks when it confidently returns stale docs. this fixes that

@dubdotco @mintlify @upstash @aisdk @claudeai Dub's support was already solid, curious to see what Pedro brings to the table. Smart hire 👏

@dubdotco @mintlify @upstash @aisdk @claudeai Support automation rising.

@dubdotco @mintlify @upstash @aisdk @claudeai This stack is honestly elite. Mintlify docs are so clean they basically reduce support tickets on their own. Curious how you're chunking content for the vector search though — that's where most AI support breaks down

nice. keep cooking!!

@dubdotco @mintlify @upstash @aisdk @claudeai AI support done right is a massive moat. most tools treat it as an afterthought — you guys are actually building it into the product DNA

@upstash @dubdotco @mintlify @aisdk @claudeai Why not mintlify mcp?

@dubdotco @mintlify @upstash @aisdk @claudeai The big UX win here is alignment: show the user what the model thinks they asked, where it pulled facts from, and a clean handoff to docs or a human when confidence is low. Curious if you are logging real support outcomes yet.

@dubdotco @mintlify @upstash @aisdk @claudeai ah yes, the "magic formula" that every AI startup discovers independently every 3 months spoiler: it's just RAG with extra steps and a prettier UI

@dubdotco @mintlify @upstash @aisdk @claudeai the RAG + human handoff combo is underrated. i built something similar for a side project and the biggest unlock was letting the AI say "idk let me get a human" instead of hallucinating an answer. kills trust instantly when it guesses wrong

@dubdotco @mintlify @upstash @aisdk @claudeai Very cool!!🚀

@dubdotco @mintlify @upstash @aisdk @claudeai i am curious how big chunk of users use that, are you able to share that? 🧐

@dubdotco @mintlify @upstash @aisdk @claudeai Good question — we still need to add o11y around it but will report back when we have some solid numbers

