These are the top 10 posts of 2025. I like to do a roundup each year as a retrospective of how much the industry has changed through some of these posts from the beginning of the year. Some are outdated already & some presaged accelerating trends.

AI Drove the Largest New Bookings of Any New Product : In January, ServiceNow announced that their AI product was the fastest growing product in their history. Since then, many other companies have reported massive growth in their AI businesses. It was a harbinger of the tectonic shift to come.

Things that Used to be Impossible, but are Now Really Hard : Kevin Scott’s quote attracted people’s attention and this post expanded on the idea that AI will both produce significant technical advances but change the way that we work by comparing the performance of an account executive with and without AI.

Theory Two : Theory is growing! I was thrilled to share the news & express gratitude to our team & broader community.

What If LLMs Change the Business Model of the Internet? In the beginning of the year, we spent time understanding whether data acquisition for training and inference of large language models would obviate the need for advertisements. There’s a significant bid-ask spread between data buyers and data sellers that remains, but many of the large language model companies have started to license data. However, in the end, the advertising model is very likely to persist as a way of commercializing the consumer base. Those subscriptions will dominate in the enterprise.

Higher Levels of Abstraction : AI is revolutionized computer programming. Andrej Karpathy tweeted about using development environment using AI. I did the same after his tweet using it to automate some of the statistical analysis that I do. I was stunned of the level of statistical sophistication, which highlighted to me how much I’ve forgotten since my graduate statistics work.

Why Databases Won’t Charge for Storage in the Future : The Tabular acquisition earlier this year fundamentally changed the way that databases are thought of. Historically, compute and storage, in other words, the calculations and the data were bundled together, but now with new file formats like Iceberg and most recently Amazon’s launch of their S3 metadata store, there will be a separation between the query engine and the storage layer. Enterprises prefer this architecture both for costs and control and an ability to build applications and data pipelines on top of a common data store.

AI Advantage for Startups: Changing the Workflow through Services : The last 10 years have been dominated by selling new workflow systems. AI changes all that. The primary sales mechanism today is selling services as a way of changing internal processes to ultimately develop a new system of record, which produces a long-term moat.

The Market Size Mistake : Tobi Lutke, CEO of Shopify, tweeted about how many venture capitalists, including me, underestimated the market size for companies and over the last 17 years being a venture capitalist. I think this is the most consistent mistake we make : startups create market size rather than a market size being a constant.

The Sudden Repricing of Startups in Early 2024 : In February, we published research showing that AI companies trade at three times non-AI companies in the public markets. We built an AI public index that continues to show AI companies in the public markets trade at higher multiples because they project approximately 50% higher growth rates. The private markets parallel the publics.

The Art of Product Management in the Fog of AI : Product management has been reinvented in the way of AI because AI is non-deterministic and chaotic. Unlike building a classic software product where a click produces the same result every time when a user engages with an AI product, the result is unpredictable or far less predictable than it has been in the past. So product management needs to evolve to work with these kinds of systems.

Thank you to everyone for reading and writing in!