Buffett Letter, More AI

Berkshire Hathaway Annual Letter

Warren Buffett published his annual letter over the weekend. You can read it here. Here are some comments:

• Allocation. Both Google and ChatGPT got this wrong so I thought it would be helpful to fill in the details. Berkshire holds $334B in cash(treasuries), $295B in stocks, and the rest in 189 wholly owned operating businesses. Given the market cap of $1.07T, that means the 189 ‘private company’ portfolio is worth about $440B. Another way to say this is Berkshire is 31% in short term bonds, 28% in stocks and 41% in private businesses.
• Cash commentary. Journalists are quick to read Buffett’s mind (a fool’s errand) by saying the high cash levels represent his bearish view of the market. Berkshire is 69% allocated to American equities, hardly bearish. Buffett is a careful allocator and the cash balance at Berkshire represents patience. It’s a reminder that it’s OK to let existing holdings run and let cash build up over many years until the right opportunity presents itself.
• Apple commentary. Apple is 25% of the public portfolio but represents just 6.7% of the Berkshire market cap overall. This unintentionally and ironically matches the S&P 500 weighting of 7.3%.
• Non-US. Berkshire owns 5% of BYD as of July 2024, the Chinese car maker which dramatically undercuts pretty much all global car manufacturers. Berkshire also owns stakes in five Japanese conglomerates that got a special shoutout in the annual letter. Buffett has always focused on the US economy as the main contributor to his success. That doesn’t mean he shuns non-US business entirely. Remember that the late Munger was an Alibaba shareholder. Every rule has a reasonable exception, even for someone as quotable as Warren Buffett.

Khan Academy ChatGPT – Tutor Me

ChatGPT has an app store that’s not all that popular but occasionally has interesting use cases. Khan Academy has one called Tutor Me which is unbelievably cool. You can try it here. Ask it to teach you about differential equations and ask it to test your knowledge. We are in an incredible golden age for self motivated learners.

Vibe Coding

We may be only a year away until “vibe coding” becomes a core part of the computing experience for the average user. If you’re unfamiliar with that phrase, it refers to describing what you want a computer to build for you via speech. The AI will write an entire application based on your description. Cursor.com is the current vibe coding leader. Replit.com is a good way to experience this if you are completely new to coding. Of course, the issue with this overall is you’re trusting the AI to do everything perfectly, which we know it doesn’t do. I can imagine in a year or two that Apple Intelligence or its SDK will allow you to ‘vibe code’ right onto your laptop and other Apple devices.

AI Etymology

One of the apps I’ve kept on my phone through the years is an etymology reference app to look up the origins of various words. Here is an AI driven Etymology app built by a talented 18 yr old from Berkley, CA.

The cool thing about using AI over a reference library is that it can essentially estimate the etymology for words that don’t exist, like company names:

• Enron – A place overgrown with thorny bushes (which happens to be perfect for hiding accounting fraud)
• Palantir – Something that can see over vast distances, shining from afar (this is a Lord of the Rings reference based on Tolkien’s Elvish language, but this description happens to work for a defense intelligence company. That’s basically what it means in Elvish)
• NVidia – Envy with a vision (appropriate for a stock that’s increased 2000% over 5 years!)