Alternative Data Weekly #269
Theme: The data supply chain needs to be watched closely.
Special thanks to our sponsor Maiden Century.
Ingest alternative data effortlessly with clean, ready-to-model flat files designed for quant workflows, dashboards, and custom research - a bright addition to your season; say hello@maidencentury.com.
QUOTES
“Without serious observability, your agents are effectively flying blind — you can’t debug them, prove value, or stay compliant”. – Ben Lorica
News
Pods
Charts
Final Thoughts (AI Artistry)
#1 – Asymmetrix published Bright Data reaches $300m ARR amidst demand for web data - Asymmetrix Newsletter #94. December 2025.
My Take: Increasingly, winners in the race to create winning AI products (i.e. those products that work, gain trust, and create obvious value) will be those products that have the best input data.
The “data supply chain” will come into focus, where high-quality upstream data is now recognized as fundamental infrastructure.
#2 – Ben Lorica published Are Your AI Agents Flying Blind in Production? December 2025.
My Take: This “quality” issue is going to get a lot of attention in 2026. As the author states, this observability is a prerequisite, not an after-the-fact add-on.
This article does a very nice job laying out the why (hopefully this is obvious!?) and the how (offline, online, RTFD ... “real-time fault detection”).
That said, these are hard problems. Is your system going to send you 47 notifications, but ignore the root cause? Will trust in the system break?
#3 – Daniel Goldberg published a few AlternativeData and AI trends that I believe will accelerate in 2026. December 2026.
My Take: Good list here. The author’s four (4) predictions hit on a few consistent themes of the ADW. A deep understanding of your workflow is key. Consumption-based pricing will be a thing. Private company data will be in demand.
BONUS: Matt Ober published 2025 Review ---> 2026 Predictions. December 2025. “Companies that own their data get more and more appreciation and respect from investors. We have seen hundreds of AI for investing or AI for financial services. The winners will start to emerge in 2026, and owning your own data will be a big part of this.”
What else am I reading:
Andriy Burkov published Artificial Intelligence #306. December 2025.
Alexander Campbell published 25 for 2025: Year in Review. December 2025.
Datactionmentor published No one cares about my data pipeline. December 2025.
Cindy Lin published Closing the Year with Reflection and AI Insights. December 2025.
Billy Newport published The Walled Gardens Are Back (And This Time It’s Your Data). December 2025.
Jordan Hauer’s published three key takeaways from Bloomberg’s The State of Alternative Data. December 2025.
Daniel Petzold published Data Milestoning: The Key to Faster Queries, Lower Costs, and Smarter Historical Data. December 2025.
David Brunner published Evals Are Forever: The Bottleneck in Agentic AI for Finance. December 2025.
Rylan Chase published What Is AI Trading? How Artificial Intelligence Trades Markets. December 2025.
Naveen Ramu published Top 10 AI Data Collection Companies in 2025. November 2025.
Source: Tim Osborne of Monte Carlo published What is AI Observability?. December 2025.
My Take: This is a very dense 8-minute podcast. He covers A LOT. What is AI Observability? The goal is to minimize downtime & maximize value. The key is to broaden the scope of what you are looking at to determine IF something breaks, and WHY something breaks.
What is success? How to define good?
What is AI observability monitoring? 4 things:
Response accuracy (how close to true)
Explainability (how this response came to be)
Model performance
Cost
Two features integral to any AI observability solution:
1- Evaluations (using AI to monitor AI)
2- Tracing (similar to lineage for a data product) 3:34…AKA explainability
Does AI observability make AI more reliable? (no, not really). Iceberg analogy … there is a lot more to it below the surface. What masquerades as a model issue is actually a data issue in disguise (the iceberg stuff below the surface).
Common problems
1. Poor source data quality
2. Drift in the embedding space
3. Confused or incomplete context
4. Output sensitivity
5. Prompt changes
6. Many, many more things that can go wrong in an AI pipeline that have nothing to do with the model output…
Need to look at 4 things:
1. Data
2. System
3. Code
4. Model
“Observability” for AI means you are observing the entire thing … not just one step in the system (most solutions just look at #4).
Highlights (8-minute run time)
Minute 01:00 – intro
Minute 02:20 – what is success?
Minute 03:00 – critical features to AI observability
Minute 04:15 – Does AI observability make AI more reliable?
Minute 05:10 – common problems
Minute 06:00 – data and AI are interdependent
Minute 06:40 – look beyond the model (4 things)
Minute 07:15 – can you resolve the issue (not just identify it)
Source: Patrick Swain and Chris Gormley of Blue Street Data published Defining the Data Economy. December 2025.
“Data has become a critical asset, but its value remains difficult to measure. This paper confronts this challenge by developing a composite valuation framework to estimate the total economic value of the U.S. data economy.”
JF: trying a new way of showing these charts … “gallery view”:
BONUS: Ben Lorica’s Are Your AI Agents Flying Blind in Production? December 2025.
AI Artistry.
I have an ongoing debate with my aspiring musician daughter about the impact of AI.
Sources: Ayesha Bhatti published History Shows Why Creators Should Embrace AI, Not Fear It. August 2025.
Kai Williams published An AI “tsunami” is coming for Hollywood — here’s how artists are responding. December 2025.
Rick Beato published I’m Sorry...This New Artist Completely Sucks. August 2025.
Prof Tom Yeh published Can AI Find its Own Poetic Voice? December 2025.
We both agree it is amazing technology, and the output is wildly good (perhaps not great).
She is fearful that there is no soul and true artists will be relegated to the background.
I do not think technology will replace human ingenuity, imagination, and artistry.
That said, this new technology will revolutionize the way art is created. This is yet another tool to help us make more and better art.
Optimism will win the day.
Human artists can certainly do better than this:
“Grok - give me an image of musicians cowering in fear from AI-generated music and feeling pessimistic about the future while not realizing this is going to be a tool that makes it much easier to express yourself as an artist.”
