If you need an investor that can put on their big boy pants and lead your round, call me.

pivot might be the most overused word in startups. half the time it's a founder following the business to its next logical step, which takes much more foresight and skill than the word gives it credit for

So much gets mislabeled as a “pivot” in startups. But as Arjun Malhotra (@BadCapitalVC) explains, Meesho’s journey wasn’t random experimentation. It was a natural evolution of the business model. “People say Meesho (@Meesho_Official) pivoted from this to that. But really, it’s like backward integration in a traditional business. Once you gain power over suppliers and move up the value chain, that’s exactly what’s happened over here.” “Vidit (@viditaatrey) has just been incredibly foresighted about knowing where value can be created next and being able to move to those things.”

Zepto filed its DRHP this week. A few interesting bits: - Ad revenue grew from ₹49cr to ₹1,636cr in two years. Qcomm is going towards the same tension as Amazon. Once brands pay for placement, what ranks highest is partly about who paid, not just what's best for the buyer. - Marketing cost per order fell from ₹33.75 to roughly ₹1 last quarter. The money draining out is going into capacity they've built ahead of demand. And demand catching up to capacity is a far easier problem to have than capacity chasing demand. - Orders per store per day went from 1,325 to 2,140. Most of a dark store's cost is fixed, so the blended numbers should improve on their own as the store base ages - Median delivery time went up, from 10.6 to 12.3 mins. Slower deliveries usually mean batched ones, which usually mean cheaper ones. This could be an attempt at cost discipline ahead of the IPO

I mess around building little things with AI pretty often and it's been time well spent. Mostly it keeps me close to how the tech is actually moving, which is hard to get from just demo videos or threads. And there's some nice downstream stuff - I think I evaluate deals a bit better, and our internal structure is now way more organised. Also, the execution has gotten cheap enough that there's little reason not to.

agentic AI is incredible at producing vitamins at scale, and that's not a bad thing. plenty of category-defining products started as vitamins. netflix wasn't urgent at launch, qcomm wasn't either - both graduated eventually to being a painkiller. the whole point of the painkiller framing was that pull comes from the product earning a place in someone's routine. that was always the hard part, and it hasn't gotten easier just because execution did

What's really the future of AI? Are models supposed to replace software or make a lot more of it? This came up in the last @benedictevans presentation, i think it depends on what we mean by software. If software means the standalone app with its own UI you log into, models are replacing it. Once the chat surface does the work, you stop opening the app. If software means the plumbing; payments, ledgers, identity, models continue to make more of it. That layer gets to stay simple and precise, because the model handles everything messy that used to sit on top. And if software means the judgment work; the drafts, decisions, summaries, models are building on a category that didn't exist before. Five years ago this wasn't software at all as it was just someone's job.

What always struck us most about Meesho was never the numbers. It was how much they cared about their sellers, how much time they spent with them, how much the team always wanted to learn from them. It's so gratifying to see that a decade later, those same sellers have a formal path to public markets through the platform they helped build.

3 frontier companies heading to public markets within months - spacex, openai, anthropic. you almost never see three category-defining companies file this close together. the last time we saw this shape was the late 90s. netscape went public in 95, yahoo in 96, amazon in 97. that wave ended up rewiring almost everything. how we shop, how we read, how we talk to each other, what a large scale tech company even is. this set is pointed at something much larger. one is trying to make humans multiplanetary, and two are building the thing that might do most of the thinking. so the first cluser IPO wave made information cheap to move while this one is making intelligence cheap to make. we tend to only recognise these moments looking back. it's strange to be standing inside one while it's still being priced.

.@deepigoyal said that blinkit has a better shot than amazon at winning qcomm. his reasoning is about the order you build things in - > amazon started with a huge catalog & slow delivery, and has spent years trying to make that delivery faster. > blinkit did it the other way, started with 10 min delivery and is now widening the catalog. delhi's assortment went from around 35k to to 80k in a year. goyal thinks adding selection onto speed is the easier path than adding speed onto selection. i think he's right. what's interesting is what that says about the burn. both companies spent enormous money to get here, but they were buying different things. amazon's money went into being the cheapest & widest, and the moat was that you learned to open amazon first because it had everything it was at a good price. blinkit's money is buying speed, and speed is a different kind of moat. once you can get something in 10 mins, the thing that decides which app you open is who has it in stock right now. you're ordering more often in smaller AOV with a delivery charge on every order, so a discount on the items rarely tips the decision. the bet i agree with is that speed-first is the harder thing to copy, and the harder thing to copy is usually the better place to have started.

we're investors in a company that uses AI for construction. the default is that the on-site contractor makes the decision about what's to be done when, based largely on anecdotal experience. this invariably leads to delays because construction is such a complex procedure. the team built an AI that breaks down the project into a defined work plan with tasks & labour requirement. now, the contractor/supervisor just needs to get the work done, instead of figuring out what to do. AI became a force multiplier - it allowed people to take on what they're best at. this is just one example.

insane stuff. the anthropic line was effectively flat for 2 years, then did a 5x in < 6 months.

Receiving goodies has been my favourite part of being back in India and being on X ✨ Thanks @GoodCapitalVC @BadCapitalVC

whats the best tool for giving your openclaw a way to make phone calls?

every elder care pitch i see is built around either loneliness or utility. both work, they just work at different points of the funnel. loneliness is the impulse. an NRI kid at 2am will respond to "your dad is alone" in a way no feature list can match. but by the second invoice, it's about utility. the buyer's looking at what got done - appointments booked, medicines refilled, ACs fixed. what gets the signup and what keeps it are two different sales, even when it's the same buyer.

in case you missed getting the physical copy of our annual book, we've made the digital version publicly available for everyone - india-technology-review-2025…

Solar Square, Orange Health, Nuuk, Entri, Xhipment. Different industries, completely different business models. The path from pre-product to meaningful revenue scale? Nearly identical. 1/14

Who wants to cofound a cross-border D2C mango business with me?

@aiweekendsxyz build-station btw thanks to @shreyansalecha @BadCapitalVC for such a cute office major missing @0xratnakar & @tushaarmehtaa

Fantastic event, great hosts, and such a pretty office! Thanks again @0xratnakar and @tushaarmehtaa for the initiative, @xshivank and @not_acp for hosting, and @shreyansalecha @BadCapitalVC for the venue! Looking forward to many more such events!