Alternative Data Weekly #226
February 21, 2025
Thanks for being here!
The theme that emerged in this week’s email is … putting data into context enhances value.
QUOTES
“… a well-designed business strategy is singular and defines how an organization wins in the market. If a company wins with the help of data and AI, these strategic choices are simply part of the business strategy, forming an integrated set of choices that create a compelling logic for how the organization can sustainably outperform the competition.” – Jens Linden
News Articles
Podcasts
Cool Charts
Final Thoughts (Context is King)
#1 – Jens Linden published Data vs. Business Strategy. February 2025.
My Take: I liked this article. Some of the ways the author suggests data can be leveraged include: Control, Automation, Decision making, and Innovation. This, of course, is specific to a business and/or individual’s context. A common theme across articles like this one is the importance of company culture and keeping the focus on solving business problems.
#2 – Alexandra McLeod and Jay Long published Culture eats code for breakfast: Rethinking AI strategy for banks. February 2025.
My Take: We’ve seen AI projects go from “this is interesting” to “show me ROI”. If AI projects are going to deliver high value, we need to get beyond crafting emails and summarizing papers. Foundations need to be built with a focus on data quality & creating a culture in which people are comfortable with data. Identify the right problems to solve in your specific context (there is that context word again). Get a couple of small wins and build momentum. This will result in a lasting competitive advantage.
#3 – Cimba.ai published Future of Data Products. January 2024.
My Take: Good overview of current data products & their shortcomings. The next generation of data products will automate intricate data processes & eliminate the need for manual intervention, streamlining workflow and accelerating business responses. Data quality is highlighted as a barrier. Good data quality leads to trust. Trust leads to more use.
BONUS: Intercontinental Exchange and Reddit Collaborate to Create and Distribute Data Products for Capital Markets. February 2025. “We look forward to leveraging our deep experience with alternative datasets to create products and services utilizing Reddit’s Data API that can help give our customers valuable insights into the markets and also help them manage risk across their portfolios.” - Chris Edmonds, President of Fixed Income and Data Services at Intercontinental Exchange
What else I am reading:
Caminao published Knowledge-driven Decision-making (Part 1). February 2025.
Tristan Handy published A Year of Innovation in AI (Part 1 of 2). February 2025.
Boris Liberman & George Danenhauer published How to Apply Alt Data Best Practices to AI Systems. October 2024.
Sara Brown published You don’t have to wait for generative AI transformations. February 2025.
Dave Hannibal published On-Premises vs. Cloud: Navigating Options for Secure Enterprise GenAI. February 2025.
Dhruva Bansal & Nihit Desai published To Reason or Not to Reason: Is 5% more accuracy worth >5x cost?. February 2025.
Timothy B Lee published Tesla's robotaxi strategy looks a lot like Waymo's. February 2025.
Sven Balnojan published Four Fun Books That Will Change How You Think About Data and AI. February 2025.
Source: Carolynn Levy & Jessica Livingston of The Social Radar podcast interview Scale AI founder Alexandr Wang. February 2025.
My Take: New podcast for me. I was curious to learn more about Scale.
Data is talked about as a raw material … Scale started with data from autonomous vehicles and chatbots. Really started focusing in on data for AI. Changed the name to Scale. Now focused on data labeling and powering data for AI, a move which has created billions of dollar of value.
Quotes: “Do too much” & "I've never seen ordinary effort lead to extraordinary results.”
Alexandr has discovered the importance of writing as a leadership tool.
Minute 44:
“So just like the entire AI industry is moving from chatbots to agents, from talking to doing, the same thing is happening for data. So one of the major trends, the biggest trends is really getting lots and lots of data for agents, which ends up looking a lot like how should in humans, when you go about and you're doing tasks and you're doing things, you have a thought process and then maybe you go collect some information and then you think a bit more, and then eventually you take the action. So for example, if you're going to book a flight, you're going to first check out what the options are, and then you're going to figure out what your constraints are and figure out, oh, can I leave at that time? You'll check your calendar and they'll also figure out, Hey, what are all the constraints?”
“That kind of chain of thought and that kind of chain of activity doesn't exist anywhere right now.”
“None of this data is out there. And so a huge trend for us or a huge thing that we're focused on is how do we actually go out and build the mechanisms to generate a lot of this data, which is a combination of how do you capture a lot of this data from just the things that people are naturally doing, as well as how do you build software systems that are able to generate this kind of data? So that's where data is going towards more and more data about people doing things and what their thought processes are when doing things. We kind of refer to this all as agent data.”
Minute 49:
“…has it ended up being something that myself and the entire scale team are very proud of as being quite central to the AI industry and everything that's going to happen over the next few years.”
Highlights (61-minute run time)
Minute 01:00 – podcast starts + Alexandr’s background (from a long line of physicists).
Minute 03:30 – Gap year after high school; immersion into AI/Neural networks
Minute 08:00 – Application to YC
Minute 12:00 – Pivot from doctor app to data cleansing
Minute 18:00 – First customers
Minute 21:00 – Early investors were AI believers
Minute 24:00 – Start of LLMs (had done a lot of work on autonomous vehicles)
Minute 26:00 – D.O.D. work
Minute 28:00 – Shift to a focus on Gen AI data
Minute 31:00 – MEI
Minute 38:00 – Growth in number of employees (very hard to maintain culture)
Minute 44:00 – Where is data going?
Minute 47:00 – Human – AI symbiosis
Minute 51:20 – Rise of GPT
Minute 56:30 – Interview wraps
Source: Jue Wang, Anne Hoecker, and Chuck Whitten published What Is Agentic AI?. February 2025.
Lots of interest!
SOURCE 2: Jens Linden published Data vs. Business Strategy. February 2025.
BONUS: Source: Matt Turck (AI - please be more than just this!)
Context is King
This is from a LinkedIn post I published.
This idea of “Agent Data” or “Access to Context” has been the beginning of articulating something I have struggled to articulate about AI (and something on which we are focused at ModuleQ).
How does the AI Model or AI Agent get to know the individual so the AI Agent can help most effectively?
From my post:
I really enjoyed this Social Radars podcast conversation with Alexandr Wang of Scale AI. Thanks Jessica Livingston & @carolynn levy of Y Combinator.
Alexandr articulated something I have been wrangling with lately. He referred to it as "Agent Data". How will AI Agents know what to do for us? Won't it take longer to describe to the agent what we want done than to just do it ourselves ... I've seen Soren Larson of Crosshatch call it "Access to Context". Erik Schluntz & Barry Zhang of Anthropic discussed this at length on the back of their published "Build Effective Agents" paper in December 2024 (YouTube link in comments).
From Alexandr:
“That kind of chain of thought and that kind of chain of activity doesn't exist anywhere right now. None of this data is out there. And so a huge trend for us or a huge thing that we're focused on is how do we actually go out and build the mechanisms to generate a lot of this data, which is a combination of how do you capture a lot of this data from just the things that people are naturally doing, as well as how do you build software systems that are able to generate this kind of data?"
Erik Schluntz posits that "every pull request triggers an agent to update all of your documentation" ... over time, the system would know you very well & know what is likely the next best action for you.
We at ModuleQ understand that context and workflow is paramount to delivering the right information at the right time. Without an understanding of the individual's persona, you'll frustrate & annoy, rather than help, the human. We are seeing strong engagement with our AI Alerts as we anticipate & then, unprompted, deliver to the right information at the right time.
"The future is already here, it just isn't evenly distributed yet."