In 2013 and 2014, USV made a number of payments-related investments and we went deep on both consumer and business-to-business payments infrastructure.

By the latter part of that decade, we had come to believe that the credit card system, and the interchange fees that make up much of the cost of the credit card system, is eventually going to go away.

Since then, we have had an active investment thesis that we called The End Of Interchange.

We use Notion AI to maintain a thesis description based on past meeting notes, emails, and internal discussions related to the topic. Our current thesis description says:

USV has had a longstanding interest in the end of the interchange/credit card system and the move to stablecoins and bank-to-bank direct payments.

The credit card interchange fee model—where merchants pay 2-3% per transaction—is vulnerable to disruption from:

• Stablecoins enabling near-zero-cost value transfer

• Bank-to-bank direct payments (ACH, FedNow, etc.) with identity/security layers

• Pay by Bank platforms

• Adding cryptographic identity layers to ACH

• Will stablecoin rails win, or retrofitted ACH/bank rails?

• How do fraud and reversibility concerns get addressed?

• What's the timeline for meaningful interchange compression?

What's great about having these constantly evolving and updating thesis descriptions is that they help us monitor news and company formation activities and allow us to be proactive in reaching out to founders to get to know companies working in an area we are deeply interested in.

We also link every potential investment to one or more of these thesis areas to make sure that our investing activies are aligned with the areas we are most interested in.

We have always deeply believed in thesis-driven investing at USV. For many years, we did this in a very unstructured way based on lots of debate and discussion. With the advent of AI and the ability to add structure to unstructured data, we are moving to a more structured approach to thesis-driven investing that opens up a lot of possibilities for us. It is very exciting to me.

Every year I put together a playlist at the end of the year with some of the new music I found and got into.

A bunch of these songs are under the radar which is my favorite kind of music.

So I hope you find something new that you like in here.

A theme I have been hearing from founders lately is training an AI to create a virtual advisor. There are various flavors of this idea but here are a few I've heard in the last week:

1/ Train a virtual board. Pick ten to twenty real people (alive or dead) and collect all of their public speaking, public writing, and all that has been written about them, train an AI on them, and then ask them questions you would ask your board.

2/ Train a virtual CMO. Pick a bunch of the top CMOs and collect all of their public speaking, public writing, and all that has been written about them, train an AI on them, and then use this as your CMO.

3/ Train a virtual CFO. Pick a bunch of the top CFOs and collect all of their public speaking, public writing, and all that has been written about them, train an AI on them, and then use this as your CFO.

I would imagine there are startups out there that are building tools to make this sort of thing easier to do. I have not personally talked to any of them, but I am sure they exist.

But in all of the cases that I heard, the founder used one or more of the popular large language models (Gemini, ChatGPT, Claude, Qwen, etc)

If you don't have the advisors you need, this is one way to address that.