India's First Claude Code Event: What I Took Away
I didn't know what to expect walking into India's first Claude Code event, organized by the Claude Code Central London (CCCL) community. Events around AI tooling can sometimes feel polished but hollow. This one was different.
What struck me almost immediately was how openly people talked about things not working. There was no performance, no pretending. Builders sat around and genuinely shared where they got stuck, what blew up, what they had to throw away. That kind of honesty is rare and honestly makes the whole room more useful.
The Takeaways That Actually Stuck
Three things came up across different conversations and the panel, and they've been sitting in my head since:
- Guardrails don't limit you. They help you build something you can trust. I used to think of guardrails as constraints. After this event I see them more as the thing that lets you hand off responsibility to the system without worrying about it going sideways.
- Orchestration is not obvious, but it matters more than people think. A lot of demos look clean. The underlying orchestration that keeps things working reliably at scale is where the real work happens and most people skip thinking about it until it breaks.
- First principles are underrated. When AI systems behave unexpectedly, the instinct is to patch. The better move is to go back to basics and understand what is actually happening. Sounds obvious. Rarely practiced.
The Panel
The highlight was the panel with Rasagy Sharma, Sathish K S, and Shashwat Ghosh. They talked about building real-world AI systems, not just demos. The gap between a good demo and something that holds up in production is wide, and hearing people who have actually shipped things talk about that gap was refreshing.
I also had conversations with people working across design, data, engineering, and GTM. That mix of perspectives is something you don't get from just reading about AI online. Everyone is approaching the same technology from a completely different angle.
Leaving with More Questions Than Answers
That is actually a good sign. I came in with a rough sense of where Claude Code fits in my workflow. I left thinking harder about how I want to structure my own projects, what I actually want to automate versus keep manual, and where the real leverage points are.
Big thanks to Vikram Pawar and Vasu Kothamasu for hosting and pulling this together. Glad I could be in the room.
