Daily Brief - How the AI warned about the market drop

TLDR: Is the AI really useful? I mean, is it really useful in finance? Sure, it can help you write emails, best man speeches, and even arrange a travel itinerary for your in-laws. But if you’re hanging around trading floors, what would be really cool is for all that machinery to process the information around us and maybe give us some hints on, say, where the market is headed in the short term? Now THAT would be really useful. In finance.

It turns out, we may be close.

Above is the snapshot from the newly launched Toggle Terminal that shows you - in red shaded areas - when the information Toggle AI was processing started to turn bearish on the S&P 500. After being bullish for much of the year (green areas) it turned bearish on July 15, right at the top of the current market correction. It then signaled - correctly - that the relief rally was just that: and the market promptly turned back down.

While not a perfect crystal ball, the AI has gotten better and better at taking in information across hundreds of corporate data points (down to minute details, like an increase in debt service on a small revolving loan), macroeconomic variables (change in trends in jobless claims, for instance) and then looking for price behavior patterns around any unusual dislocations. At Toggle, we give it access to huge volumes of structured and unstructured data it’s able to tap into on a daily basis - and it goes through all of it.

Aggregated across thousands of tradable securities, this kind of analysis provides a fascinating assessment of market health based on its individual constituents. It’s also a very practical application of the new technology. In finance.

Of course, we’ll need a lot of proof before we trust market calls from an AI analyst. But compelling evidence is fast emerging. One good example of successful information harnessing with LLMs is in the area of financial analysis. A paper by Kim et al. shows there is definitely potential to improve upon Street consensus by unleashing LLMs on thousands of financial statements. The Chain-of-Thought (CoT) prompt that effectively “teaches” the model to mimic a financial analyst produces earnings forecasts that hold enormous promise based on the paper’s accuracy calculations (they outperform analysts by 8-10 percentage points).

An autonomous AI analyst is definitely coming, and we are actively working to build that capability. But even before that, the age of AI in finance has firmly changed how some investors are now harnessing information in the market to make better investment decisions.

Market Movers: What next for S&P Sectors?

Below is the historical 1-month response from S&P sectors, following previous highs in the Toggle Leading Indicator:

1. S&P Real Estate: 3.41%
2. S&P Health Care: 3.28%
3. S&P Utilities: 2.97%
4. S&P Technology: 2.85%
5. S&P Financials: 2.61%
6. S&P Telecom: 2.32%
7. S&P Consumer Discretionary: 2.30%
8. S&P Consumer Staples: 2.26%
9. S&P Energy: 1.87%
10. S&P Industrials: 1.17%
11. S&P Materials: 0.39%

Earnings Spotlight: TSMC's July Surge

Taiwan Semiconductor Manufacturing Co. (TSMC) reported a robust 45% increase in revenue for July, reaching NT$256.95 billion ($7.9 billion), significantly outpacing its growth in the previous quarter.

This strong performance, driven by sustained demand for artificial intelligence (AI) chips from major clients like Nvidia Corp., suggests that TSMC may exceed analysts' expectations for third-quarter revenue, projected at NT$747.4 billion. As a key supplier for AI accelerators and the sole provider of processors for Apple's iPhones, TSMC's impressive growth underscores its critical role in the ongoing surge in AI-related demand.

The company recently raised its full-year growth outlook, indicating confidence in continued strong performance beyond its previously guided mid-20% range.