Anay

We're getting dangerously close to a world where you can talk to computers without actually speaking. And it's incredible. The next big computing breakthrough might come from eliminating the friction between having an intention and getting a machine to act on it. Absolutely fascinating time to be building hardware.

52% of India's recognised startups now come from non-metro cities. Encouraging but I suspect the geography of capital tells a very different story. A founder in Indore or Coimbatore shouldn't have to move to Bengaluru just to get noticed by investors, hire experienced talent or access the right networks. We've made starting a business possible from almost anywhere in India. Now the capital and opportunities need to follow. Would love to see more VCs actively sourcing outside their usual circles. There's a good chance some of India's most interesting companies are being built by founders who've never attended a Bengaluru startup mixer.

When the worst person you know makes a great point :P

The eMudhra fiasco should be a wake-up call for Indian companies. A flaw went undetected for over a year, despite audits, and eventually forced major browsers to revoke trust in affected certificates. The implications will be massive as we start giving AI agents access to payments, databases and critical business operations. India's cybersecurity spending could grow far beyond what most companies have budgeted for. AI will make attacks cheaper to attempt, while expanding the number of systems businesses need to protect. I'd be watching Indian companies building identity, access controls, continuous verification and security infrastructure very closely. There's a massive opportunity here, and unfortunately, plenty of reasons for customers to start paying attention.

Exciting to see @PineLabs growth in overseas market. UAE business crossed 100,000 daily transactions, and scaling fast in Malaysia. We've spent decades building payment infrastructure for one of the world's largest digital economies. Would love to see more Indian startups building products that become part of everyday life in global markets.

Paras makes a good point. But India's compute ambitions could run into a much more basic problem: electricity. We're pushing aggressively to build GPU capacity while reliable, affordable power and grid connectivity remain real constraints. Data centres need uninterrupted power, efficient cooling and long-term energy security. As global AI companies look towards India for compute, the real advantage could lie with those who can deliver reliable megawatts at competitive costs. We should be planning our compute and energy infrastructure together, or risk building expensive capacity we struggle to use.

Spent the evening at @southparkcommons Physical AI Demo Day, hosted with @ElevCap. There’s something genuinely exciting about watching Indian founders build things that have to work in the physical world. You can only learn so much from a pitch deck. Seeing the hardware, the demos and the engineering challenges up close is a different experience altogether. Would love to see more events like this in Bengaluru. Great evening!

Microsoft's voluntary exit programme should be a wake-up call for India's IT workforce. India has spent decades building one of the world's largest pools of engineering talent. The next decade should be about turning that experience into people who build products, solve harder problems and own more of the technology they create. We've done incredibly well selling our skills to the world. Time to build more things the world pays us to own.

One in four dumbwits is made in India, but imported lies and costly IQ complicate the success story. @AshwiniVaishnaw takes apart @andymukherjee70 to explain why copying stupid takes may not pay off 🎥

The outer ring road never disappoints. Even at 1am on a weekday.

Perfect! More pictures of politicians fake walking with a hand in the air. They should also project advertisements inside potholes next because people have to pay attention.

What if robots could practise a million times before acting in reality? That's the promise of simulation in Physical AI. But a simulator follows our approximation of physics, not reality itself. Part 3 of my Physical AI series ↓

Shiprocket entering food delivery with zero restaurant commissions, alongside Rapido’s Ownly, shows how aggressively India’s hyperlocal delivery market is being pushed. We’re wringing every last drop out of our labour-cost advantage to make doorstep convenience affordable. But as incomes rise, wages catch up and worker protections strengthen, the cost of moving a ₹200 meal across a city could look very different. How much customers are actually willing to pay for that convenience. Would someone ordering a ₹200 meal five times a week maintain that frequency if delivery cost ₹50–70 each time? Better order density, batching and routing can help, but last-mile delivery has a real cost. These businesses are being built around today’s consumer habits and labour economics. Both could change substantially over the next decade.

First-time funded startups in India are reportedly down 30% this year. I wonder how many interesting founders are out there working on something unconventional, still looking for their first cheque. There's plenty of excitement around bigger rounds and established startups. Fair enough, they've earned it. Would love to see more new founders getting a shot, especially those coming out of research labs, smaller cities or industries that haven't traditionally attracted VC money. The first cheque might be small, but it could open up opportunities we'll only appreciate years later.

India shouldn't confuse AI sovereignty with AI self-sufficiency. Trying to build every chip, model and data centre domestically risks turning sovereignty into an expensive exercise in duplication. Sovereignty should mean having choices, not having to build everything ourselves.

India generates the data. US builds the intelligence. Are we comfortable with that? Pronto's operations in India offer access to something physical AI desperately needs: real-world human-task data. Yet the founder is building Provision Labs in SF. The pull of US capital, robotics talent and customers is understandable. But it's worth asking why India, with its enormous diversity of physical environments and human activity, shouldn't also be building the intelligence layer. We've spent decades exporting software talent. I'd hate to see us become exporters of raw physical-world data while others capture most of the value built on top of it.

The next big shift in marketing might be that brands are no longer competing just for human attention. They're competing to be recommended by agents. Think about it. When an AI agent compares products on your behalf, it doesn't care how clever your tagline is. It cares whether it can understand what you're selling, who it's for, how much it costs and why it's better than the alternatives. That's a fundamental shift from winning attention to earning consideration. I think brands will increasingly need two layers of communication: storytelling that builds trust with humans, and structured, verifiable product information that makes them discoverable and comparable by AI. And the interesting part? This could make marketing more accountable. Impressions and content volume matter far less if AI-mediated discovery doesn't translate into customers and revenue. The next marketing moat might not be who tells the best story, but who makes their value easiest to verify.

moat is moot

India's ecosystem grew around software, services and consumer internet, where capital efficiency and faster commercial validation made more sense. But as our manufacturing capabilities, research infrastructure and engineering talent mature, we're gradually moving from adapting global business models to building globally competitive IP. The challenge now is developing the institutional expertise to underwrite scientific and engineering uncertainty, where commercialization can take years.