For the first time, Andreessen Horowitz's Top 100 Consumer AI Apps ranks products not just by downloads and traffic, but by what U.S. consumers actually pay using YipitData's consumer spend panels. As the report shows, that 𝗻𝗲𝘄 𝗹𝗲𝗻𝘀 𝗿𝗲𝘄𝗿𝗶𝘁𝗲𝘀 𝘁𝗵𝗲 𝘀𝘁𝗼𝗿𝘆. Spend concentration is one of those perspective shifts. The top 1% of consumer AI payers spend over $10K a year on average (work usage may be even higher), and they're still growing that spend, up 80% in the last 18 months. The median payer spends $300 a year and has barely expanded it. The top 1% alone drives more spend than the bottom 50% combined, and over half of the vendors on the spend list don't appear on the traffic lists at all. Where people go and what they value are two different leaderboards. We're proud to help Andreessen Horowitz bring both into view. Read the full report from Olivia Moore: #ConsumerAI# #AlternativeData# #a16z#

We're excited to sponsor the @neudatalab New York Summer Data Summit on June 11! If you're attending, stop by Booth AN on Floor 1 to learn how investors are leveraging YipitData's growing suite of proprietary and exclusive alternative datasets to generate alpha, forecast KPIs, and build conviction across sectors. If you'd like to meet with our team during the event, schedule time with us using the link here: #alternativedata# #neudata#

We are heading to the Columbia Quant Conference this Thursday, Nov 5! At 10:25 a.m., our very own Jingwen Xu and Maxime Chapados will deliver the keynote “B2B: The Next Frontier in Alternative Data: How Yipit is Unlocking the Next Generation of Alpha Signals.” Reach out to us if you’ll be there!

Here are the current top 10 stocks on the @AltindexApp top AI stocks list Astera Labs $ALAB Anglogold $AU Granite $GVA Jack Henry $JKHY Agnico Eagle $AEM X-Energy $XE Constellation Energy $CEG Revolution Medicines $RVMD Cognyte $CGNT Descartes $DSGX Their "AI Score combines AI and Alternative data to assess a stock's mid-term market outperformance potential (6+ months) by analyzing multiple financial, technical and alternative data points. This predictive measure assigns a score from 0 to 100, representing the stock's performance and potential."

I agree with this. Classically, the Tiger cubs liked moving great analysts around to various sectors in a deep sequential immersion, and when became senior risk takers, the ability to know 3-4 sectors deeply became a huge advantage in comparing the opportunity set cross sector. The reality is each sector will have a time where it is alpha rich (feels easy) and alpha poor (feels impossible). As a healthcare investor, I know this feel SUPER well… I logic through how the next 3-5 years may play out and i think today’s multi-manager discretionary alpha will see increased competitive challenges. Day by day, the typical quarter in the life of a typical multi-manager analyst is being systematized. Quants who are much better at position sizing and execution will figure out a way to scale this, just as they have scaled alternative data signal. And the trade will have already happened before the human can press “buy”. The “super pod” idea of an elite team of perhaps 3-5 decision makers all supported by maybe 1-2 analyst per sector feels like a strategic shift that could elongate multi-manager alpha streams. Allowing slightly longer duration (which requires a bit of a mindset shift on drawdowns / idio constraints) may be necessary. Navigating crowding in this model would be key, and a nontrivial challenge.

There has been this sort of persistent "stuckness" in AI Investing where the promise of what is possible has not shown up in practical tools for investors that have meaningfully bent the curve of alpha. To me there is a very fundamental reason for this. AI can be very helpful for one off situations: let me track management commentary over 12 quarters, summarize my morning inbox, run a compensation tiers analysis with the proxy, etc. Super helpful things (if not still a bit challenging to set your system up in a seamlessly usable way). The conceptual unlock is the concept of seamlessly "connecting the dots". What does readthrough from Company A mean to revenue forecasts at Company B? How does the move in the 10-Year impact all of my companies with variable rate debt? Seamless, automated. Some of the smartest AI builders I know have been focused on this concept of company & industry ontologies. What does that mean? Better to give you an example. I was a healthcare services investor for 10+ years. The concept of medical trend was a key thing to get right. Was trend accelerating or decelerating? Would the HMOs beat or miss on MLR? Did they price their new members correctly? What is this new Part B drug launch doing to loss ratios? These questions had big implications across the healthcare service food chain. And there were various company-level and industry data sources to track (medical claims, etc), and a lot of the value in identifying these inflections came from being out in the field, talking to companies, working to identify inflections first. I made a lot of money doing this better than the average healthcare investor. If i can press a button and re-build this system, holy crap it's so exciting. I can then actually automate all of that alpha. But, not so fast. The complexity inherent in this whole system is immense, and the number of failure rates along the supply chain is huge, and can be as simple as not normalizing MLRs for an acquisition or market exit and thus having a distorted baseline to drive the model. Bulling in member files, state level reserves, the ongoing research pipe, and the clean substrate of the financial modeling corpus is a *necessary* element. As in alternative data, the first order observation is quickly priced by quants, and all the alpha is in the 2nd order observation (i.e. "UNH missed on MA MLR but it's due to their Stars reduction in this county and HUM actually shrunk there, so dumb market is selling off HUM but it's a misread and a big buy", for example.) The dept of normalizing MLRs, cleaning up the models, then connecting all of this together into one tracking system is effectively an industry and company ontology. How the economic relationships of an industry and company interact. I've talked to a number of people attacking this bottleneck in AI Investing, and they have been working to build ontologies. Sometimes, with simple examples, they work. But for complicated examples like the above, they are terrible. There is just too much complexity. And, my prior is all of the fundamental alpha is in the complicated ontologies. There is one other fundamental truth. Investors will not trust an output without interrogating the logic chain. "HUM will beat on MLR". OK, show me every input and the full logic chain as to why. This is very hard to do in python. I think it's super obvious you do this in Excel. It's the only path forward I see, honestly. Particularly to build an "over my dead body" system for institutional investors that *already* have this system built out. Let me automate that for you. "But Brett, Excel sucks and python is better". Ok, maybe someday. But navigating the cultural change of human investors needs to be considered. And, as a person who spent 10+ years working in or adjacent to quantamental strategies (with pretty good intel on what is being built out there), no one is even close to being able to building a trend ontology as described above. It won't happen in python. This is why AI Excel is the key accelerant in AI Investing, in my opinion.

AI has broadly become synonymous with LLMs, but most AI/ML models used in quantitative trading are decades old. Boosted trees and PCA still account for more alpha than anything built on transformers. More detail on several ML techniques in quant trading: Gradient boosted trees and random forests: A standard strategy in short-horizon trading is to start with an intuitive set of microstructure features and train a non-linear model on them. Random forests provide a baseline ensemble that’s hard to overfit, while gradient boosted models fit fainter structures at the risk of overfitting. Features can include microprices, characteristics of the present order book, recent returns of related instruments, momentum signals, and orderflow directionality. Pre-transformer natural language processing: Throughout the 2010s, NLP PhDs were a sought-after resource for quantitative hedge funds. A variety of techniques such as Loughran-McDonald finance-specific sentiment dictionaries, TF-IDF and bag-of-words classifiers, and LDA topic modeling were used to extract sentiment from news, earnings reports, SEC filings, and FOMC minutes. Transformers are more generalized tools for language processing but some of the older methods in the toolkit are still used for their simplicity, predictability, and cost-effectiveness. Dimensionality reduction: A statistical arbitrage researcher may have thousands of correlated instruments and only a few years of clean historical data, so the covariance matrix has more parameters than observations to estimate them from. Principal component analysis is the standard remedy. A handful of leading components explain most of the variation in a universe of equities, and hedge funds will frequently neutralize the leading factors and trade the residuals. Autoencoders generalize the same idea to nonlinear structure, but PCA persists because its output is stable, interpretable, and cheap to recompute. Computer vision: From around 2010 to 2017, hedge funds gained a large informational edge by building automated pipelines to extract alternative data from satellite imagery. The models were unremarkable by research standards, mainly derivatives of convolutional neural networks. Some of the most famous applications include counting parked cars in retail lots, gauging construction progress at industrial sites, estimating crop yields from near-infrared reflectance rather than visible light, and reading crude inventory off the shadow cast by a floating tank roof. The trade worked best during a time when imagery was scarce and the engineering was novel. Today image-based alternative data is sold by third-party vendors for relatively inexpensive consumption. The large memory footprint, slow inference, and black-box nature of LLMs make them a poor fit for alpha generation at trading firms. Hand-tuned models based on intuitional feature selection are still the norm. For niche industries like quantitative finance, there is ample room for competition with generalized frontier models.