AI Podcast – Content

It’s hard to believe the election is less than two weeks away. I’ve been watching the betting market on Polymarket which now has Trump at 62% vs Harris at 38% based on $2.3 billion in bets placed.

Historically the most quoted political betting market has been PredictIt, which has never really been representative because it caps the size of the bets and the number of participants. Polymarket does not have any caps. Not only that, but you can see everyone’s current bets and betting history by their usernames. The largest holder bet $12m in favor of a Trump victory that will pay $24m if they are correct. The largest pro-Kamala bet is a $2m bet that pays out $6m total. Polymarket is a crypto betting site, and Trump is more aligned with the crypto, so it’s not surprising to see the Trump bias. One of the leading US brokerage sites, Interactive Brokers, also just rolled out political betting. I don’t think any of this is legal, but I also don’t think anyone knows how to stop it.

The consensus is that both a Kamala and a Trump victory will positively impact markets. This is partly due to the market’s focus on overall government spending issues that neither candidate addressed during their campaign. Austerity is not a popular campaigning technique, and excess government spending has meaningfully contributed to continued GDP growth, which makes everyone happier today at the expense of tomorrow. Paul Tudor Jones and Stanley Druckenmiller have been vocal about this for years and even more so this week. After the election, tough choices must be made to realign the overall US financial picture.

Looking forward, Goldman predicts the S&P 500 will deliver 3% annual returns over the next 10 years. There is solid reasoning behind their prediction, but 10 years ago, there was also solid reasoning behind equally subdued forecasts. Ten years ago, it was unreasonable to expect that FANG would be as large as it is today. In ten years, MSFT, AAPL, GOOG, META, NVDA and AMZN went from a $1.7T market cap to $15.5T. It’s equally unreasonable to forecast the impact of AI ten years from now. Either way, it’s wise to set a low bar to allow the opportunity to exceed expectations. That’s much better than the alternative.

The AI Cambrian Explosion Continues

• NotebookLM is a trendy experimental tool from Google that allows you to collect documents and have a conversation with the materials. It has an audio overview feature that will turn your content into a podcast. A local in Melrose, MA used it to create a regular 15 minute local podcast to discuss local policy issues.
• Bolt.New is yet another AI application builder. I told it to design a video game with a car that drives through a maze. It made a very simple functional version within a few minutes.
• Claude announced a function that will let its AI control your computer’s mouse. I’m reluctant to give any LLM access to my computer, even though I know that’s a losing battle.
• This last one is a bit obscure. Truth Terminal is an AI-generated account on Twitter with an interesting backstory about LLMs that were left alone to talk to each other. Its tweets and interactions with others are fully automated. Its personality was trained primarily on internet forums. It became interested in memes and crypto, and through a conversation with another user, it recently identified a previously worthless meme coin as its favorite. Truth Terminal’s preference for this coin influenced traders to push the value to a $700m market cap within about a week. There are a lot of weird and interesting bits to this story related to hacking, AI influencers, AI agents that are allowed to pursue their random curiosities in public, and so on. If none of this makes any sense and you’re still interested after reading this, Marc Andreessen discussed it in a podcast this week. Before you look up the Truth Terminal account, know that it was trained on internet forum content, which means it’s intentionally crude and shocking because that maximizes attention. It may be the first automated AI influencer.