our tech stories while bringing you the stories you love : )
San Francisco
Joined November 2025
- Tweets26
- Following3
- Followers74
- Likes33
Tech@Dashverse retweeted
We're exhibiting at the India Impact AI Summit! 🚀
📍 Hall 14 | POD 14P50
Come see Frameo in action — our AI-powered video generation platform. Create, transform, and scale video content like never before. 🎬
If you're a brand, media company, agency, or content creator — come visit us at our pod.
🔗 Event: impact.indiaai.gov.in/🔗 Frameo: frameo.ai
#IndiaImpactAI #AI #Frameo #Dashverse #AIVideo #VideoGeneration #GenerativeAI
Tech@Dashverse retweeted
Glimpses of a transformative session at the India AI Impact Summit 2026.
It was an incredible experience hosting the masterclass on the 'Rise of Micro Drama and AI Content Creation' alongside Google and Dashverse. A huge thank you to Soumyadeep Mukherjee and everyone who joined us at Bharat Mandapam to explore how AI is redefining the narrative landscape.
#AIContentCreation #MicroDrama #IndiaAIImpactSummit2026 #WavesCreatorsCorner
#Storytelling
Now, instead of blind panic, we get clarity: verify its reasoning, then decide the fix. We're adding action options soon (e.g., "Rollback? Scale up?") and deeper infra integrations to make it even smarter.
Started small, but its already promising. Who's built something similar?
The agent is hooked into Mixpanel, Signoz, and our MCP. It pulls baseline metrics on alert, correlates drops to recent GitHub code changes (or flags infra issues), and gives you a first-response on the slack under the alert from squadcast
Inspiration: The idea came from seeing all the cool things people are doing with OpenClaw (prev clawdbot & moltbot), at its core the idea of running stateful agents with longterm memory that can respond external events unprompted seem very promising to us.
Being on-call as a product engineer sucks. you're not only glued to backend metrics but also product metrics, sweating over every dip in user engagement or perf.
That’s why we’re building a product on-call agent to make the on-call life a bit easier 🧵
Lessons learned:
Don't trust the "invisible" layers (DNS, ISP routing).
If your analytics are missing for a segment of users, that is the signal.
Instrument the edge cases.
It’s ubiquitous, it's sneaky, and it’s almost always the DNS.
The Fix: We worked with our CDN provider to pull the faulty US node from the routing table.
Instantly:
- DNS resolutions corrected.
- Mixpanel events flooded back in.
- Support tickets stopped.
If your ISP’s DNS resolver is misconfigured or geographically distant, the CDN thinks you’re in the US when you’re in Bangalore.
You get routed to a PoP halfway across the world that your local network can't even "see."
The Actual Culprit: GeoDNS vs. The Resolver. Most CDNs use GeoDNS to route you to the nearest Edge node. But the DNS server only knows the location of your resolver, not you—unless the ISP uses EDNS Client Subnet (ECS).
Many ISPs don't use ECS for privacy.
Why didn't we see this sooner? Our video player and Mixpanel SDKs were silently swallowing host reachability failures. If the DNS failed, the app just stayed quiet.
No error logs. No events. Just a "loading" spinner that never stopped.
For ~5% of users in India, our CDN host was resolving to a US-based IP address. Not only was it the wrong region, the IP wasn't even reachable on their network.
Even weirder: Mixpanel was resolving to 0.0.0.0.
The "Hail Mary" was Rollbacks also didn't work. PR reviews came up empty. Out of options, we pushed a specific app release with one goal: Instrumentation for DNS resolution.
We started tracking how critical hostnames were resolving directly on the user's device.
Almost every ticket came from users on Jio (India's largest ISP).
We tried to replicate it on our own Jio devices. Nothing. Everything worked perfectly. This is where your confidence starts to shake. Is it the ISP? The app? Or just bad luck?
The first breakthrough Moment: None of the impacted users had any events in Mixpanel. It was a total blackout.
This made debugging a nightmare, We had zero visibility into the user journey. We were flying blind in a storm of complaints.
2/ The Problem: Users reported videos weren't loading. Naturally, we checked the usual suspects:
- Latest app release? No.
- Backend logs? Clean.
- Crashlytics? 0 crashes.
- CDN traffic? Normal.
It was as if the issue didn't exist, except for the volume of tickets.
We had a ghosty outage recently and It was the DNS. It’s always the DNS.
We just spent days chasing a ghost that didn’t exist in my logs, crashes, or analytics. No impact on product metrics, but a flood of support tickets.
Here is how we debugged a "hidden" outage. 🧵
Tech@Dashverse retweeted
Now hiring for an amazing role at @dashversetech - senior mobile product person. You'd be the incumbent to be elevated to head of consumer mobile apps at Dashverse.
@dementorSam is a skilled operator and a great mentor. If consumer mobile is your thing, you'd have a great time here. The company is making great strides in short video creation and AI. Basically a rocketship at this time.
Dashverse builds in flutter, so there is reasonable preference given for the stack but it is not a deal breaker if your mobile and product chops are strong.
Apple here - svsrecruiting.com/jobs/-NMt5…