The following analysis was written independently by Tuesday Capital and has not been authored or commissioned by Context Analytics or BridgeWise.
New academic research validates the edge of BridgeWise and Context Analytics combination.
A study published last month by Runjing Lu (University of Toronto), Luka Vulicevic (UC San Diego), and Yongxin Xu (Monash University), titled Do LLMs Make Markets More Efficient?, delivers the first causal evidence on a question the finance industry has been asking for two years: do AI tools like ChatGPT and Claude actually change how markets price news? Previous work on the topic has been largely descriptive, or based on controlled experiments that cannot say what AI is doing in live markets. This study uses a natural experiment, reports effect sizes that are economically meaningful rather than marginal, and is the first to isolate the impact of AI availability itself on price discovery.
Here is how they did it. Every so often, OpenAI, Anthropic, and Google have outages. Servers crash, software breaks, updates fail.
These outages happen roughly every four days on average and have nothing to do with markets. So the researchers compared two kinds of days side by side: days when news hit while AI tools were
working, and days when news hit while AI tools were broken. The finding: when AI is available, prices
finish reacting to news faster. Specifically, when a news story is released, most of the price move happens that same day. But a smaller, predictable piece of the reaction normally spills into the next trading day.
When AI is available, that next-day piece shrinks by roughly 46% to 61%. This means markets are reacting more quickly with LLMs.
The paper also shows that investors who can keep processing news during AI outages earn roughly two to three times the typical news-strategy return in those windows.
WHAT THIS MEANS FOR THE VALUE OF CA
Three points from the paper map directly onto the structural advantages that sustain the acquisition thesis.
First, CA does not depend on OpenAI, Anthropic, or Google staying online. The paper shows that when the major AI providers go down, a large share of the market temporarily loses the ability to process news, and a simple long-short strategy on news sentiment earns on average two to three times its normal
return (the return the same strategy earns on days when AI is working) during those windows. In other words, the investors who can still trade on news during an outage capture returns that everyone else
temporarily cannot.
The paper identifies those investors as ones with "proprietary or locally deployed models that remain available during outages at major providers." CA fits that description directly: the CA pipeline runs
on proprietary infrastructure with in-house NLP that continues delivering signals when LLMs are down. Independence from the major AI providers is a clear economic advantage, and we expect this to support
CA's competitive position as text processing becomes more central to how markets price information.
Second, the profitable window for news-based signals is short, which makes real-time delivery essential. The paper shows that most of the post-news price correction is completed by the second trading day,
and the gap between outage and non-outage returns converges by that point. A signal delivered in a daily report or an end-of-day file misses the window in which the return actually exists. CA delivers sentiment metrics every minute through a real-time JSON API. The tweet-level feed scores individual posts within 300 milliseconds of receipt from X. The S-Factor feed refreshes every minute with
fifteen distinct metrics per security, benchmarked against a 20-day baseline. This cadence matches the horizon in which the paper shows the returns live. Competitors operating on slower delivery models are increasingly disadvantaged baseline.
Third, CA processes data that competitors cannot access. The company holds a commercial license from X (formerly Twitter) to use its content for financial analytics, one of very few in the world. CA also processes Reddit across 50+ subreddits, StockTwits, financial news from 4,000+ sources, SEC and global filings, earnings call transcripts, and podcasts. A hedge fund using a general-purpose LLM to measure sentiment on a specific stock gets a summary pulled from whatever news articles and public posts the model's search tool happens to find, which is the same summary any competitor using the same LLM would get. That is not an edge. CA clients receive something different: the full conversation across licensed sources, quantified, scored against each security's own historical baseline, and delivered as a
number they can trade on. As LLM commoditization accelerates, the relative value of exclusive data rights rises rather than falls.
OTHER ADVANTAGES WORTH NOTING
Twenty years of historical data. Machine Readable Filings, structured corpus of every SEC filing back to 2006 parsed section by section, enabling strategies based on how company disclosures evolve over
time. CA's word-count-change strategy using this data has generated roughly 10.5% annualized returns on the short leg versus 7.2% for the control group.
Three U.S. patents covering the extraction, evaluation, and calculation pipeline. Broad coverage across roughly 4,500 US equities, 2,200 ETFs, 1,600 European names, 800 ASX stocks, 800 crypto tickers, and 500+ private companies, all delivered through real-time JSON APIs.
CONCLUSION
The study is the first causal evidence that AI-driven text processing has become central to how equity markets price news. The three advantages laid out above, running independently of the major AI
providers, delivering signals fast enough to capture a short return window, and holding data rights competitors cannot replicate, are what separate the data providers that benefit from this shift from those that do not.
Context Analytics meets all three, having spent over a decade building precisely the business the paper now describes. The findings support the investment thesis behind Bridgewise's acquisition of CA and
reinforce our conviction in the trajectory of the asset over the hold period.
To have full access to the research, click here.
investors@tuesdaycapital.com.br
tuesdaycapital.com.br
Av. Dr. Cardoso de Melo, 1.308 conj 22 - Vila Olímpia - SP / BR
