Invest with AI

A year or so ago, dropping more than five PDFs into ChatGPT was enough to make it lose the thread. Now tens of thousands of expert transcripts can run through an MCP connection, and Third Bridge was one of the first expert networks to set that up. On this episode of Invest with AI, Brett sits down with Dan Thomas, who runs partnerships at Third Bridge. Dan started there ten years ago as a content analyst on European high yield and distressed credit, covering everything from nickel and zinc smelters to Mediterranean ferry companies, and has his name on more than a thousand expert interviews.

“The principal consumer of the data is going to be an application. It’s going to be agentic, and they’re going to be fairly unsentimental around what the front end actually looks like.” – Dan Thomas

Getting there meant reworking a transcript library that goes back to 2011 and was never tagged with a machine in mind. That work is now part of the buying decision, as funds start asking vendors what it costs to run their data through a model. And if expert libraries do turn into beta, Dan expects funds to spend more on senior, harder-to-reach experts that a library can’t replace.

Watch the full episode to hear what Third Bridge had to rebuild to get there.

TIMESTAMPS
0:00 Intro
0:44 Nickel Smelters, Ferry Companies and 1,000 Interviews
4:34 Who Writes the Transcripts
7:03 Third Bridge's Early Move on MCP
9:10 Your Next Reader Is an Agent
11:45 Rebuilding a Transcript Library for Agents
16:27 Can an LLM Hear the Hedging?
18:56 Click-Through Rates on Citations Keep Falling
22:22 “We Now Have 45 MCPs”
26:05 How Funds Test an MCP Before Buying
28:10 Backtests Still Want the Flat File
30:41 Experts Taking Calls From AI Interviewers
33:34 There’s Only One Expert
40:12 Why PMs and Risk Teams Want Expert Content Now
44:05 Mapping Suppliers and Customers Back to Your Thesis
48:24 Are Expert Libraries Becoming Beta?

Explore more episodes of Invest with AI:
https://www.fundamentedge.com/invest-with-ai

Want to up-level your investing style? Learn more at Fundamental Edge: https://www.fundamentedge.com

AI agents are only as good as the real-time data they reason on, and that shift is about to change how alternative data is bought, priced, and used on Wall Street. On this episode of Invest with AI, Brett sits down with CarbonArc founder Kirk McKeown, who spent two decades building the primary research and data operations at Glenview and Point72 before going out on his own.

Kirk walks through how credit card panels went from real alpha to expensive table stakes ("paying alpha prices for smart beta"), why he built CarbonArc around consumption pricing and MCPs that plug straight into AI agents, and how markets could trade as agents get plugged into live data.

TIMESTAMPS
0:00 Intro
0:35 From Tudor to Glenview to Point72
2:43 Industrializing the Channel Check
8:33 The Craft of a Great Expert Call
11:08 Be a Counterparty, Not a Service Provider
13:33 How Alt Data Rewrote Consumer Research
16:34 Where Alpha Goes When Data Goes Mainstream
20:40 Turning Raw Data Into a Real Signal
27:21 How Card Data Distorted Price Discovery
30:09 Why He Started CarbonArc
33:26 Working With (and Around) the Data Incumbents
39:43 AI Is Only as Good as Its Inputs
40:41 Will AI Change How Markets Trade?
44:09 Can Claude Be Trained to Pick Stocks?
45:29 How to Use CarbonArc

Explore more episodes of Invest with AI:
https://www.fundamentedge.com/invest-with-ai

Want to up-level your investing style? Learn more at Fundamental Edge: https://www.fundamentedge.com

Former MIT professor and current Fundamental Research Labs CEO Robert Yang wants to build a network of agents that operates like a hedge fund, with analysts researching companies and a portfolio manager coordinating the work. After managing 50 agents at once across seven monitors (which he called a terrible user experience), his goal is to give investors one agent that coordinates the others, with a system they can adapt to their own process.

Robert and Brett discuss where investment research is headed as AI takes on more of the analyst’s work, how investors’ roles could change, and why Brett is joining FRL to help build it.

Timestamps:
[00:00] Intro
[00:43] Why Robert Wanted to Build AI Before It Was Popular
[08:40] Running an MIT Lab and Questioning the Work
[14:05] The Agent Research That Pushed Him to Leave
[19:20] Project Sid: Building a Society of 1,000 Agents
[22:18] The Search Party Nobody Programmed
[26:58] From Minecraft to Excel Agents for Wall Street
[33:05] Going Viral Wasn’t Enough: Making Shortcut Work
[37:28] Why Hedge Fund Analysts Became the Customer
[41:51] From an Excel Agent to an Investment Team
[45:51] Why Nobody Wants to Manage 50 Agent Threads
[50:12] How Close Are We to a Digital Hedge Fund Analyst?
[53:52] The Problem with AI That Always Agrees
[56:50] Teaching an Agent to Hold an Investment View
[1:04:29] When Analysts Start Working More Like PMs
[1:06:29] Brett’s Role: Training the Digital Analyst
[1:09:07] If AI Frees Up Your Time, Where Does It Go?
[1:11:34] What This Means for FE and Invest with AI

Explore more episodes of Invest with AI:
https://www.fundamentedge.com/invest-with-ai

Want to up-level your investing style? Learn more at Fundamental Edge: https://www.fundamentedge.com

Can AI replace traditional expert networks? And what happens to investment alpha when AI makes primary research faster, cheaper, and more accessible?

In this episode of Invest With AI, Fundamental Edge founder Brett Caughran and Khe Hy sit down with Sagar Kadakia, founder of Qualitate, to explore how AI-powered expert interviews are impacting fundamental investment research.

Qualitate is rethinking the traditional expert network model, using AI to conduct interviews, automate channel checks, and help investors generate proprietary insights at scale.

But as primary research becomes faster and more accessible, a bigger question emerges: Where will investors find their edge when everyone has access to better information?

IN THIS EPISODE:

- How AI-powered interviews compare to traditional expert networks
- Automating channel checks and scaling primary research
- Using AI to conduct dozens of expert interviews in days
- How investors can incorporate their own research and investment theses into AI interviews
- The role of human judgment in AI-driven research
- Compliance, MNPI, and expert authentication
- Whether AI will commoditize investment alpha
- Why asking better questions may become the next competitive advantage

READ THE FULL EPISODE RECAP

Explore the key takeaways and watch/listen to the podcast.

https://www.fundamentedge.com/invest-with-ai/invest-with-ai-episodes/12/qualitate

CONNECT WITH US

Fundamental Edge: https://www.fundamentedge.com
Invest With AI: https://www.fundamentedge.com/invest-with-ai

Subscribe for more conversations exploring the intersection of fundamental investing and artificial intelligence.

CHAPTERS (TIMESTAMPS)

00:00 Intro: Can AI Transform Expert Networks?
00:55 Introducing Sagar Kadakia and Qualitate
01:28 The Origins of Qualitate
05:02 The Breakthrough in AI Voice Technology
08:31 Human Judgment vs. AI Interviewing
14:37 Automating Channel Checks and Protecting Research
19:23 Using Existing Research to Improve AI Interviews
21:13 How Qualitate Works: From Research Question to Insights
24:44 Removing the Scheduling Bottleneck
26:32 Preventing Fraud and AI-Generated Expert Responses
29:37 MNPI and Compliance Safeguards
35:10 Quantitative Investors and the Future of Alpha
37:49 Will AI Commoditize Investment Research?
40:52 Can Expert Supply Keep Up?
43:28 Building an AI-Powered Research Library
46:53 When Investment Alpha Becomes Beta
49:05 Why Speed and Better Questions Matter
54:13 The Future of AI-to-AI Conversations
57:32 Scaling Qualitate and Building an AI Business
1:01:40 Final Thoughts: The Future of Investment Research

As AI becomes more deeply integrated into fundamental investing, where does human judgment still provide an edge?

In this episode of Invest with AI, Brett Caughran and Alistair Smallwood, Head of Applied AI at Primer, discuss how artificial intelligence is reshaping equity research, alpha generation, and the future of fundamental investing.

"And I think there'll be a real opportunity for the sharpest people in fundamental investing that really understand AI systems to identify over and under-reactions in the short term of stocks in sectors that they know really well." - Alistair Smallwood

From systematizing mosaic theory and identifying investment signals to automating earnings forecasts, AI is transforming how fundamental investors conduct research. But as these systems become more widely adopted, they could also introduce new biases, market mispricings, and opportunities for investors who think a step beyond the data.

As AI becomes more capable of replicating traditional investment research, the question is no longer just what these systems can do, but how investors can use them to gain a competitive edge. 

Watch the full conversation to hear where Brett and Alistair believe the next opportunities for alpha could emerge.

Timestamps:

00:00 — Introduction
00:27 — Alistair's Background: From Investing to Applied AI
03:08 — The Evolution of AI in Equity Research
06:54 — Building AI Systems That Adapt to New Models
08:46 — Mosaic Theory and the Future of Investment Research
11:50 — Scaling Mosaic Theory with AI
17:02 — Can AI Replicate Human Investment Research?
19:51 — The Future of AI Skills and Autonomous Research
22:38 — Evaluating AI Output in Fundamental Investing
24:10 — How AI Could Reshape Alpha Generation
27:38 — AI Bias and Potential Market Mispricings
30:44 — AI-Generated Earnings Forecasts and Market Expectations
33:00 — The Limitations of AI Benchmarks
37:07 — Where Human Judgment Still Matters
42:19 — AI's Impact on Investment Horizons
47:02 — The Future of Factors and Risk Models
48:07 — The AI Skills Gap Across Hedge Funds
50:31 — Why Fully Autonomous Investing Remains Challenging
56:09 — Finding Alpha Beyond First-Order AI Insights
57:54 — The Risks of Misinterpreting Alternative Data
59:36 — Building and Scaling Primer
1:03:36 — AI Adoption, Compliance, and Data Security
1:06:58 — The Future of AI Adoption in Investment Firms

Explore more episodes of Invest with AI:
https://www.fundamentedge.com/invest-with-ai

Want to up-level your investing style? Learn more at Fundamental Edge: https://www.fundamentedge.com

It was never worth an associate's time to strip every KPI out of every CIM that crosses the desk. At $20 of tokens, it is. No guest this week. Brett Caughran and Khe Hy on the work that just became economic, and why the hard part is no longer the technology.

The Mac Minis people bought to run OpenClaw mostly run Codex now, and the reason is reach: sessions on your machine are reachable from your phone. No MCP for DealCloud? The agent drives your logged-in Chrome instead. Which moves the constraint. Both of them train investors on this, and both say the question changed two months ago, from how do I use these tools to what is worth building.

Brett is unsentimental about where that goes wrong. An agent turned loose for 24 hours gives you a long chain of mediocre work and a large token bill. The back half gets into what does work: the five layers, the 20/60/20 primer rebuild, why an out-of-the-box primer hands you Zacks and Motley Fool, and the headless RMS that goes and gets the research sitting in OneNote, Slack, Bloomberg IB, and Outlook.

-----------------------------------------------

Timestamps:

[00:00] Intro
[01:13] — Grokbot, Cursor, and a Three Horse Race Again
[02:56] — Are the Mac Minis Mothballed?
[04:05] — Why the Power Users Moved to Codex
[06:22] — The Always On Machine You Run From Your Phone
[07:27] — No MCP? Let the Agent Drive Your Browser
[09:44] — What This Means for Wall Street's Claude Fluency
[10:49] — Training Shifts From "How" to the Art of the Possible
[13:05] — Brett on the 24 Hour Agent Slop Chain
[14:08] — The Faster Horse Era of AI
[15:00] — $20 of Tokens vs. a $100K Associate
[16:14] — Why AI Makes Lazy Research Easier
[17:24] — Five Layers, and Why Decision Support Is the One
[18:27] — From Answer Engines to Work Mechanisms
[20:38] — The Notion Problem: Staring at a Blinking Cursor
[22:21] — Every Investor Wants Something Different
[23:55] — Rebuilding the Primer Skill: The 20/60/20
[26:36] — Excel Fluent Models and the Boat Pricing Tracker
[28:02] — State of Play: MCP for Investment Firms
[29:57] — Why Out of the Box AI Hands You Zacks and Motley Fool
[31:42] — BlueMatrix, Third Bridge, and AlphaSense's Walled Garden
[32:46] — The Headless RMS and Where Research Actually Lives
[33:50] — The 16 Column Limit Nobody Documented
[37:12] — What a Real Research Dashboard Looks Like
[38:34] — Markdown Extractors as a DIY Knowledge Graph
[41:59] — The Always On Earnings Preview
[42:46] — The Codex Moment for Public Equity AI
[45:53] — Three Pillars of the Midsummer Inflection

-----------------------------------------------

Want to actually build these workflows yourself?
The AI Accelerator is Fundamental Edge's 6-month cohort for investors who want repeatable AI workflows. Learn More below:
https://www.fundamentedge.com/ai-accelerator

Watch the full podcast series on our site: https://www.fundamentedge.com/invest-with-ai

Follow Invest with AI on:
Spotify: https://open.spotify.com/show/033xcEEovVViS7hIYwNuGZ
Apple Podcasts: https://podcasts.apple.com/us/podcast/invest-with-ai/id1896918892

Doug Clinton started with one question in late 2022: can ChatGPT beat the S&P 500? The early results were promising enough that he built a company around it. Today he runs Intelligent Alpha, where frontier models do the investment analysis and the portfolio management and he's still a partner at Deepwater, making the same calls as a human.

He, Brett, and Khe get into where the models are already good enough, why he grades them a B+ analyst and not an A, and the contrarian call on which model is actually best for stock work.

A conversation on AI for hedge funds and the buy-side: knowledge graphs, ontologies, model routing, obsolescence risk, and what it takes to build an AI-native investment process.

-----------------------------------------------

Timestamps:

[00:00] Intro
[02:39] — Can ChatGPT Beat the S&P 500?
[04:11] — Why LLMs Keep Handing You $NVDA
[05:05] — Prompt Engineering to Agentic Workflows
[06:45] — Organizational Context: Every Portfolio You've Ever Built
[09:07] — Public Markets Alpha Is a Power Law Game
[10:09] — Where Human Intuition Still Beats the Model
[11:14] — The Hybrid: Part Quant Book, Part Fundamental Book
[12:30] — The Three Buckets of Data
[13:58] — Can Agents Do Channel Checks?
[14:49] — Is MCP Institutional Grade Yet?
[16:11] — Grading the Models: A B+ Analyst
[16:55] — Knowledge Graphs Are the Frontier Right Now
[18:41] — Schemas, Ontologies, and the Gray Matter
[19:21] — Taste: Getting AI From B+ to A
[20:26] — What Building an EBITDA Ontology Actually Looks Like
[24:04] — The Dev Shop Wants to Interview Your PM
[25:37] — Trade Perfection for Speed
[26:36] — We've Built a Lot of Things We No Longer Use
[28:25] — The Vendor Cycle and Controlling Your Own Destiny
[30:23] — Model Routing and the IA 500 Benchmark
[32:36] — The Contrarian Take: GPT Over Claude for Stock Work
[33:53] — Codex vs. Co-work and a Branding Problem
[37:32] — Why There's No Harvey or Rogo for Investing Yet
[40:45] — Do Funds Want the iPhone or the Android?
[43:13] — You Need Someone in Leadership Who's AI-Pilled
[45:19] — Fable Built a 10-Stock Portfolio. It's All AI.
[48:28] — Brett Updates His Skepticism
[49:09] — Advice for Building an Asset Manager With AI
[51:33] — Getting Comfortable With Uncertainty
[53:03] — Go Get Lost in the Models

-----------------------------------------------

Want to actually build these workflows yourself?
The AI Accelerator is Fundamental Edge's 6-month cohort for investors who want repeatable AI workflows. Learn More below:
https://www.fundamentedge.com/ai-accelerator

Watch the full podcast series on our site: https://www.fundamentedge.com/invest-with-ai

Follow Invest with AI on:
Spotify: https://open.spotify.com/show/033xcEEovVViS7hIYwNuGZ
Apple Podcasts: https://podcasts.apple.com/us/podcast/invest-with-ai/id1896918892

Former Point72 TMT analyst Thomas Li left the buyside to build Daloopa in 2019 with the goal of capturing market share from the data oligopoly of Bloomberg, FactSet and Cap IQ. Last year data consumption on the platform grew over 100x, and it wasn't the models that did it.

He, Brett, and Khe get into what separates a reliable MCP from a brittle one, why Claude in Excel still can't handle a real buyside model, and the one layer of the analyst job he thinks is already collapsing.

-----------------------------------------------

Timestamps:

[00:00] Intro
[01:29] — Founding Daloopa in 2019 to Take On the Data Oligopoly
[02:29] — Accuracy, Latency, Trust: Why Human Data Collection Breaks
[03:54] — Coach-Built vs. Production Line: The Data Factory Analogy
[05:45] — From 5% of the Database Used to Almost All of It
[09:22] — The Real Unlock: Chatbot to Connected MCP
[10:12] — Not All MCPs Are Created Equal
[14:25] — Where MCP Reliability Is Actually Improving
[16:16] — The Cannibalization Debate: Should All Your Data Go in the MCP?
[17:54] — Building Your Stack: Which Pipe Do You Turn On?
[21:02] — One Pipe or Nine? Where Multi-Vendor Stacks Hallucinate
[24:31] — The Moat Is the Factory, Not the Tool
[26:59] — Going Deeper, and the 13-Week Problem
[31:30] — How Funds Are Building Their Orchestration Layer
[33:29] — Knowledge Graphs: Folders, Markdown, and Pre-Done Analysis
[35:44] — Open Weights vs. Frontier: Where Post-Training Wins
[40:08] — Post-Training Is an Objective Function Problem
[42:49] — If Knowledge Work Isn't Verifiable, Is There a Ceiling?
[45:28] — Can AI Pass Judgment? The Factor Investing Case
[51:40] — The Race to the Middle: Quants and Fundamentals Converge
[53:18] — Why Claude in Excel Still Can't Handle Operating Leverage
[58:47] — Eval-First Product Development
[1:01:17] — Spring 2027: Hiring, Training, and the Collapsing Middle
[1:04:45] — Coding Got Solved. Engineering Didn't.

-----------------------------------------------

Want to actually build these workflows yourself?
The AI Accelerator is Fundamental Edge's 6-month cohort for investors who want repeatable AI workflows. Learn More below:
https://www.fundamentedge.com/ai-accelerator

Watch the full podcast series on our site: https://www.fundamentedge.com/invest-with-ai

Follow Invest with AI on:
Spotify: https://open.spotify.com/show/033xcEEovVViS7hIYwNuGZ
Apple Podcasts: https://podcasts.apple.com/us/podcast/invest-with-ai/id1896918892

Kris Bennatti and Hudson Labs are now giving a no-hallucination guarantee, something she says no other financial AI offers. The reason it's rare is the uncomfortable part: her team found state-of-the-art models return the wrong financial number about 30% of the time once you go past a few reporting periods.

Kris, Brett and Khe get into why precision is still a difficult problem, how a forensic risk score turns fraud signals into a hard number, and why she thinks 2026 is "integrate or die."

"I believe we're the only people on the market that are willing to do a no-hallucination guarantee, which does give you a bit of a sense of how much risk there is in the average financial AI software." — Kris Bennatti, CEO of Hudson Labs 

-----------------------------------------------

Timestamps:

[00:00] Intro
[00:40] Meet Kris Bennatti, CEO of Hudson Labs
[01:22] From Pre-LLM Filings to Founding Hudson Labs
[04:16] Why Precision Is Finance AI's Hardest Problem
[05:35] The Test: Wrong Numbers 30% of the Time
[06:29] Launching the No-Hallucination Guarantee
[10:20] Why NotebookLM Feels Better (It's the Search)
[12:01] Vector Databases vs. Just Using Claude
[13:36] Screening for the "Most Stressed-Out CEOs"
[14:16] Have MCP and Connectors Fixed Accuracy?
[19:50] From Prompts to Skills
[20:50] MCP vs. CLI
[23:04] Integrate or Die: The 2026 Business Model
[26:52] Where Generalist Models Catch Up (and Where They Won't)
[33:36] Inside the Forensic Risk Score
[38:47] The Track Record: Score 70+, a 1-in-3 Chance of SEC Action
[44:17] The Cost Problem: $100 for Hudson, $13K on Opus
[46:26] Fable, Co-work, and the New Token Economics
[51:20] The Guidance LLMs Still Miss (It's the Verb Tense)
[54:05] Where to Find Kris & Hudson Labs

-----------------------------------------------

Want to actually build these workflows yourself?
The AI Accelerator is Fundamental Edge's 6-month cohort for investors who want repeatable AI workflows. Learn More below:
https://www.fundamentedge.com/ai-accelerator

Watch the full podcast series on our site: https://www.fundamentedge.com/invest-with-ai

Follow Invest with AI on:
Spotify: https://open.spotify.com/show/033xcEEovVViS7hIYwNuGZ
Apple Podcasts: https://podcasts.apple.com/us/podcast/invest-with-ai/id1896918892

Fable just shipped with the promise of being Anthropic’s most capable release yet. However, daily power users Brett Caughran (Fundamental Edge) and Khe Hy (Rad Reads) still can't conclude whether models have gotten meaningfully better in the last six months.

We run a round-the-horn vibe check: the eval where Fable read 10,000 minutes of Khe's calls and surfaced the one thing his business was avoiding, why complex Excel modeling still breaks, the Karpathy "knowledge wiki" for a 500-document deal folder, and the quiet case for Cowork over Claude Code.

Want to actually build these workflows yourself?
The AI Accelerator is Fundamental Edge's 6-month cohort for investors who want repeatable AI workflows. Learn More below:
https://www.fundamentedge.com/ai-accelerator

-----------------------------------------------

Timestamps:

[00:00] Intro
[00:57] — First Impressions of Fable (and Why the Ban Was a Relief)
[03:05] — The Eval That Found the Thing He Was Avoiding
[05:14] — Why Evals Are Hard, and the One-Shot Short Signal
[06:56] — The 100-Company Earnings Preview Test
[07:46] — The Bridgewater Judgment Layer (High-40s → Mid-70s)
[10:48] — Braintrust and Scoring MCP Vendors
[12:30] — The Karpathy Knowledge Wiki for 500-Doc Folders
[18:20] — Making Your File System Legible to Agents
[21:11] — Turning a 120-Page Manual into Skills
[22:04] — Thin vs. Thick Skills and Composable Subagents
[25:36] — Why Complex Excel Still Breaks (and the Subagent Fix)
[27:24] — Claude Code vs. Cowork: The Line Is Collapsing
[29:57] — The Infosec Case for Cowork (VM vs. Root Access)
[33:50] — Vibe Check: Skepticism Is Up, and "Bot-Sitting"
[36:26] — Digital Twins and Where the Buy-Side Actually Is

-----------------------------------------------

Watch the full podcast series on our site: 
https://www.fundamentedge.com/invest-with-ai

Follow Invest with AI on:
Spotify: https://open.spotify.com/show/033xcEEovVViS7hIYwNuGZ
Apple Podcasts: https://podcasts.apple.com/us/podcast/invest-with-ai/id1896918892