I flew to Bangalore for YC Startup School on April 18th.

I came back with a diary full of contacts, a clearer sense of what the Indian startup ecosystem actually looks like from the inside, and one thought I have not been able to stop thinking about since the flight home.

This post is about that thought. The rest is context.

The Event

The main event was at White Feather in Bangalore. It started with lunch and open networking, which is where most of the useful conversations actually happened for me. The structure of the afternoon was talks from YC founders and YC representatives, mostly about where AI is going and what it means for people building things right now. The talks were good. The networking was better.

Then dinner, and after dinner the room shifted into something harder to describe. People were pitching. Not formally, not in a designated slot, but continuously, to whoever was nearby. YC members were in the room. VCs were in the room. The density of ambition per square meter was noticeably higher than any room I have been in before.

The after party was organised by Staaake, and it was better than the main event.

I do not mean that as criticism of the main event. The main event was well-run and the content was worth being there for. But the after party had a different quality to it. The conversations were less guarded. People were pitching ideas that were clearly not ready, and nobody cared, because the point was the conversation. VCs were talking to founders were talking to people who were neither, all in the same room at the same time, and the permeable nature of that felt important.

I did not get the merch. I did not get the goodies. I came back with something that takes longer to explain.

The Mindset Shift

Here is what I did not expect.

My research work over the past months has been deeply concentrated. Yang-Mills extensions, BRST cohomology, protocol translation middleware, persistent homology on correlation matrices. Each of these is a narrow problem with specific technical requirements. The mode of thinking that deep technical work demands is inherently minimalist: reduce the problem to its essential structure, work within exact constraints, and do not introduce complexity you cannot justify.

That mode of thinking is correct for the work it serves.

But I had not noticed how much it had started to shape my thinking outside the work as well.

I was applying a research mindset to ambition itself. Minimise scope. Reduce to what is tractable. Do not overclaim. Work within what the evidence currently supports.

Those are the right standards for a theorem. They are the wrong standards for what you want to build.

What I saw in that room in Bangalore was the opposite approach made visible. People were not starting from what was tractable. They were starting from the largest version of what they wanted and working backwards to what needed to be true. The goal came first. The constraints were things to solve, not things to accept.

The lesson I kept returning to in the conversations and after: if you start low, you get satisfied easily.

The moment you hit the first version of the thing you set out to build, there is an enormous pull toward calling it done. If the goal you set was modest, that pull becomes a stopping point rather than a waystation. The ambition you start with determines the ceiling you are willing to tolerate.

This is not the same as being unrealistic. The founders in that room were not delusional about what they were building. They understood constraints better than most people. But they kept the largest version of the goal in view precisely so that partial progress remained partial rather than complete.

Start high. Not because realism is wrong, but because where you start determines where you are willing to stop.

The Ecosystem

I had been thinking about the Indian startup ecosystem for a while, but mostly from the outside and mostly through the lens of the defence work I have been doing. The middleware project sits at the intersection of software and defence procurement, and understanding how the indigenisation push actually moves through the system requires understanding the investor and founder side of it, not just the technical side.

What I got from the conversations at White Feather and especially at the after party was a much more textured picture than I had before.

A few things that were consistent across a lot of different conversations:

The AI layer is compressing timelines in ways that were not true eighteen months ago. Things that would have required a large team to build are now buildable by two or three people with the right approach. This is changing what is fundable, what is viable, and what the bar for a first product actually is.

The gap between what gets built and what gets used is still wide, and the people who are closing it are the ones who spent time understanding the distribution problem before the product problem. This came up in multiple forms across multiple conversations.

The room between genuine deep tech and feature-wrapped AI is getting harder to occupy. Investors are getting better at distinguishing them and founders are getting more honest about which category they are in.

None of this is secret information. But hearing it from people who are inside the system, in a room where you can immediately ask follow-up questions and watch how someone responds when pressed, is different from reading it.

What I Am Doing With This

The honest answer is that I am still processing it. The research work continues. The Yang-Mills project, the defence middleware, the TFI work. None of that changes. The technical requirements of those projects are what they are and they do not bend to changes in mindset.

But I have also been building two things in parallel alongside the research, and that is where the Bangalore trip landed differently than it might have a year ago.

When you are only doing research, events like this feel adjacent to the main work. Useful, interesting, worth attending, but not directly load-bearing. When you are also trying to build something, they feel different. The conversations are not just informative, they are calibrating. You are not just collecting ideas, you are checking your own thinking against people who are further along the path you are on.

The mindset shift I described above, the start high observation, hit differently in that context. I have been running both the projects alongside some of the most technically demanding research I have done, and the temptation to set modest goals for the projects because the research already demands so much has been real. It is the same pull toward tractability that I described earlier, applied to ambition rather than to theorems. The room in Bangalore was a useful correction to that.

The after-party reminded me not to treat the larger vision as aspirational. It is not aspirational. It is the specification.

I did not get the merch or the goodies. I did get the development pack with credits across a range of tools, which is genuinely useful when you are building and researching simultaneously and watching resource costs carefully. I came back with a diary full of contacts, a clearer picture of an ecosystem I care about, and the simple and direct observation that the ceiling you set for yourself early is the hardest constraint you will spend the rest of the work fighting. Start high.

Related: What Talking to Startup Founders Changed About How I Think About Research