Investing

Answer engines are the first consumer product to take real search share in two decades. I use it daily and watched it move from novelty to habit for people around me.

The lab behind the Kimi models, one of the few Chinese labs that ships frontier-class open weights. A hedge on the idea that the AI race stays a two-country race.

Alphabet spin-out that builds large quantitative models for drug discovery, materials, navigation and post-quantum cryptography. A bet that AI applied to physics and chemistry is the next wave after language.

When I bought, Anthropic sat in second or third place in AI and was priced far below the leader. The pitch was safer, more reliable models for business. I build on these models every day at work, so I had a read on quality that the price did not yet reflect.

What it taught meWorking in the field is an edge in private markets. I knew the model quality months before the valuation caught up.

The category leader in generative AI, with Microsoft behind it and an IPO on the horizon. I bought in two rounds and later trimmed the position to a size I can hold without watching it.

What it taught meTrimming a winner is fine when it lets me hold the rest without anxiety. Position size drives behaviour more than conviction does.

Search pays for the research, and the research is now the product. My rule on it is simple: buy below four times sales, start trimming between six and eight.

A business that funds its own expansion, with return on capital to match. Buy zone is six to eight times sales; past twelve I trim rather than add.

As AI compounds, humanity will push into milestones that never existed before, and space is one of them. Blue Origin builds on the path SpaceX blazed at a real discount to SpaceX on a sales basis. Bezos funding, NASA and Space Force as anchor customers, and an engine business that sells to competitors.

The one company that sets the technical and regulatory path the whole launch industry rides on. Reusable rockets, launch cadence and Starlink all compound on each other. I bought early and held through every mark.

What it taught meA long lockup is a feature. It stopped me from selling into the first rerating, and the rerating I would have sold into was the small one.

Inference-only accelerator chips aimed at Nvidia for LLM serving. Training gets the headlines; serving is where the volume goes. A small bet that the inference layer fragments.

Capital back if no chip stock closes below the barrier at maturity, leveraged upside if the basket rises, capped. A defined-risk way to hold the semiconductor supply chain for fifteen months.

Every company training or serving a model rents the same silicon. I hold it knowing the growth is real and that it is somebody else’s capex, which is the part that can turn.

Inference chips built around deterministic, low-latency serving. I took the position when fast inference was the obvious bottleneck for agents, and exited when the position was bought out.

What it taught meA quick exit on an AI hardware name taught me that in this sector the buyer often shows up before the product proves itself. Take the exit.

India’s gas trading exchange, promoted by the power exchange. A third exchange in the record, on the same thesis: as India moves to gas, trading volume follows.

Chennai drone maker for agriculture, defence and industry, heading for an IPO. I worked on autonomous drones at Dronistics, an EPFL spin-off, so I can judge the hardware.

Press-on nails, growing fast from a tiny base. Early, thin on cash, and a good test of whether the fund can pick beauty.

Chewable functional beverage from a studio-backed founder team. Added in 2026; I have no read yet beyond the first update.

Dehydrated pet food for a market that is only now forming in India. Same studio as Noklu, added at the same time.

India’s dominant stock exchange, still unlisted. An exchange is a toll booth on a growing economy. The IPO timing depends on the regulator, so patience is the whole trade.

India’s agri-commodity exchange. Same toll-booth logic as NSE, applied to a market that formalises as farm incomes grow.

Budget hotel aggregator that survived its own hype cycle and moved toward profit. A pre-IPO bet on Indian travel demand.

Pune company behind AeroGCS, ground-control software used by around half of Indian drone makers. Software is the part of the drone stack with margins, and I spoke to the founder before buying.

Luggage brand with licensed characters. Distribution-led consumer play. Cash is tight and the next round decides it.

Innerwear brand that reached its first profitable quarter after cuts. The odds are against it. I keep it on the list because losers belong in the record too.

India got cheap data before it got rich, and two operators are left to collect on it. In a duopoly they can finally raise tariffs, with the subscriber base still moving up the ladder.

Lends where the big banks will not: used commercial vehicles and small operators. Decades of loss data on exactly that borrower is the moat, and credit cost is the number that decides it.

The default first car for a middle class that is still forming, with the service network to match. The risk I accept is being late to electric.

The best return on equity and assets in the Indian banking market, and a bank that runs on process rather than on one person. The post-merger digestion is the thing to watch.

Diagnostics and oncology in one company, priced like a utility and funded like a research lab. Patent cliffs are the recurring risk and the recurring discount.

A pure-play pharma after spinning off the rest, which makes the pipeline the whole argument. I hold it for the dividend and the discipline, not for a breakout.

Brands people buy without thinking, in currencies that are not all the same. Slow, boring, and the kind of position I can forget I own.

Prediction markets became a real information source during the 2024 elections. The NYSE parent took a strategic stake and the company is returning to the US with a licence. Markets on everything is a product I want to exist.

The one European fintech that reached profit at scale and keeps adding countries and products. I pay for it as a customer, which is my first filter for any consumer name.

Secondary shares bought when employees were looking for liquidity. Unreal Engine is the toolchain for games, film and simulation, and Fortnite is a platform rather than a game. Unreal Editor for Fortnite looked very promising at the time.

Rolled over from the Beekeeper exit into the merged company, where I now work as Senior AI Engineer. A larger employee-experience platform with a clear plan to grow profitably, and I am one of the people building the AI part of that plan.

The online business school where I did my MBA. Mobile-first, low-cost, selective, and I know the product from the inside as a graduate.

Employee shares in the company I spent six years building. I saw the product, the customers and the numbers from the inside. Exited when Beekeeper merged with Lumapps and rolled part of the proceeds into the combined company.

What it taught meThe only private position where I had real information. Buying into the company I worked for was the easiest decision on this list, and it turned out to be the best one.

A high coupon if none of the three falls through the barrier. If one does, I end up owning it at a deep discount. All three are on my buy list anyway, so both outcomes suit me.

A holding company I can value on book rather than on a story. Buy close to book, sell somewhere near twice book, and let the float do the work in between.