Tech-Thoughts

Here’s a simple project success heuristic. All projects have some element of uncertainty but over time the questions should become more refined. If that’s not happening, then the project is almost certainly failing. The company might still succeed via a pivot but the project is unlikely to. Consider the progression of questions for a consumer (B2C) product. Would anyone use this? Would a million people use this? What’s the total addressable market? Would people pay for this? Is 10$/month too high or can I increase the price further? Notice how questions keep becoming nuanced. ...

American government goes everywhere exporting its brand of democracy. The Chinese government just wants to do business and make money [with these autocracies] in Asia, Africa, and South America - Lee Kuan Yew In many ways, American companies take the same approach. They build one product, localize it a bit for the target countries and try to win big there. However, Chinese companies seem to take a totally different approach. Utilizing their lower labor costs than their American counterparts, they build targeted products for markets. And Temu is the perfect example of that. Temu’s parent company’s flagship product is Pinduoduo. Pinduoduo is designed for the Chinese market. And a completely unrelated E-commerce product, Temu is designed for the American market. ...

Javascript isn’t a language. It is just a specification. Well, that’s true for many other languages like Python and Java. However, most languages have just one de facto implementation. For example, CPython for Python. Java has two major ones - Oracle’s JDK and Google’s Android Java. That’s not true for Javascript though.

There are two major ways to make money on consumer Internet - subscription or ads. Also, unlike enterprise-world, usage-based subscription is hard and confusing for consumers. So, we end up with 10$ a month all-you-can-consume buffet or we end up with ads. Now, like it or not, not all users are created equal. Some are more valuable than others. And these valuable users indirectly pay more than the users with less spending power. In other words, these users subsidize the product for the less fortunate fellows. In fact, in some ways, ads are a transfer of wealth from rich to poor users of the same service. So, next time wonder when people complain about ads, they will always be rich and powerful. These rich users would prefer a minuscule 10-a-month price instead of being worth thousands, if not millions, for the advertisers. And if the product has network effects, like social media, then it is unavoidable for these richer folks. ...

Bengaluru-based Postman , a API management product, has raised a total of $433 million valuing it at ~$5.6B. Its revenue was $171M in 2023. As a comparison, Rubrik with ~$850M ARR is valued at $6B or about 7 times its revenue. The same metric would put Postman’s valuation at $1.2B. I would argue further that Postman has much lower stickiness and will face higher churn and hence, the revenue multiple would be lower for Postman. ...

For more than 10 years, this website has been running on WordPress. Over time I come to dislike it for multiple reasons.

The true value proposition of LLMs lies in their ability to convert unstructured data from sources like websites and documents into structured information with reasonably high accuracy. Yet, the real profit lies in the products built on top of LLM technology. Each year, approximately 4 million books are published worldwide. On average, a book contains fewer than 120,000 words, translating to less than 160,000 tokens in LLM (Large Language Model) terms. Imagine if every single one of these books were generated by GPT-4—it would amount to an astounding 640 billion tokens. At $5 per million tokens , generating all these books would tally up to about $3.2 million! Let’s say the book market represents only about 1% of the total LLM text generation opportunity. Even then the total addressable market of LLM text generation is approximately $300 million annually—a modest figure when compared to AWS, which raked in $90 billion in 2023 as the cloud market leader.

Generated code, ideally, should not be committed to version control. Committing generated code can sometimes speed up testing and code generation but it is a design smell. It is better to cache generated code via CI caching . Committing generated code to version control is the worst as it is hard to even detect the difference. However, there are a few specific circumstances where committing generated code/config/data to version control is worth it. ...

A few years back, I scraped data on podcasters from iTunes. The data was a bit underwhelming and made me realize that podcasters can’t be a potential market. It is a bit dated but I believe is still relevant.

Two cryptocurrency exchanges came out early on from Y Combinator. One is 2012. One in 2013. One returned 1500X to early investors. The other one ceased to exist after 2 years. What happened?