Financial research with Claude
Summary
Speaker Background
Brooker Belcourt brings a unique non-engineer perspective to AI-assisted financial research. His background spans a decade at hedge funds doing investing work, followed by a decade in fintech startups. Most recently, he ran the finance vertical at Perplexity before transitioning to consulting work helping investment firms adopt AI tools.
Core Thesis
Claude Code represents a paradigm shift for financial research because it removes the compute limitations of web-based LLM interfaces. While ChatGPT and similar web apps cap processing time at approximately 20-30 minutes, Claude Code running locally can process for much longer periods, enabling the creation of comprehensive, interactive financial dashboards that would previously require hours of manual analyst work.
Key Demonstration
Brooker demonstrated creating an earnings preview dashboard for Meta with a single command. The system:
- Accessed financial data via the Daloopa MCP
- Read local files including company transcripts and notes
- Called relevant APIs
- Generated an interactive Streamlit dashboard with multiple tabs
- Produced dynamic charts based on company-specific guidance metrics
The entire dashboard generation took approximately 20 minutes of compute time but replaced what would have been 5+ hours of manual analyst work.
Revolutionary Approach
Instead of producing static outputs (Word documents, PowerPoints, PDFs, or emails), this approach generates interactive, live dashboards. The system is flexible enough to adapt to each company's unique guidance patterns - Meta guides revenue, expenses, CapEx, and tax rate, while other companies might guide only EPS or revenue.
Prompt Management Strategy
A critical insight is managing prompts through GitHub repositories rather than web interfaces:
- Problem: ChatGPT's 8,000 character limit constrains prompt complexity
- Solution: Store prompts as Claude plugins in version-controlled GitHub repos
- Benefit: Prompts can be much more robust, include investment philosophy, data source priorities, and specific instructions
Investment Philosophy Integration
Brooker emphasizes teaching LLMs your investment philosophy because they tend toward consensus thinking. His approach includes:
- Prioritizing trajectory and accelerating businesses
- Specifying preferred data sources
- Defining what constitutes "great investing"
- Setting analysis preferences (beat/miss track records, guidance analysis, etc.)
Data Sources Mentioned
- Institutional: Daloopa MCP for robust financial data
- Retail: Perplexity Finance for transcripts and financials
- API Access: Fiscal AI for programmatic data access
- News: X (Twitter) for real-time financial news
The Compute Gap Opportunity
The session highlights a significant opportunity gap: since early 2025, LLMs have been capable of much longer processing times, but web applications haven't kept pace. Claude Code exploits this gap, explaining its recent surge in attention among power users.
Key Takeaway
Building a library of prompts as intellectual property in GitHub is extremely valuable. These prompts, combined with Claude Code's extended compute capabilities and MCP integrations, allow individual analysts to build Bloomberg-like dashboard systems customized to their specific research needs and investment philosophy.
Key Concepts
Claude as Research Assistant (Not Just App Builder)
While most discussions focus on Claude building applications, Brooker emphasizes its power as a research assistant that can access MCPs, local files, and APIs to produce comprehensive financial analysis.
Notable Quotes
"I'm not an engineer. My background is about a decade working at hedge funds, investing, and then a decade in fintech startups."
"I was at Perplexity running the finance vertical and have since been doing a lot of consulting and catching up investment firms in this world of AI."
"A lot of it is Claude is building apps and that's really cool, but I found it an incredibly powerful research assistant."
"With Cloud Code, it's really powerful because you can access MCPs, you can access all the files on your computer, and with one command, you can actually create this dashboard."
Tools Mentioned
Transcript
BROOKER BELCOURT (Every Consulting) - Financial research with Claude
=== BROOKER BELCOURT (Every Consulting) - Financial research with Claude ===
(07:29:36): See ya.
(07:29:36): See ya.
(07:29:38): All right, guys, we're going to close out the day.
(07:29:42): It's getting dark out in New York,
(07:29:44): and I don't want to say we're saving the best for last,
(07:29:48): but I'm a big fan of Brooker.
(07:29:49): So excited to have him share some of the work that he's been doing.
(07:29:54): And Brooker, I'll turn it over to you.
(07:29:58): Hey, everyone.
(07:29:58): It's great to be here.
(07:30:00): Just some quick background on who I am.
(07:30:02): So I'm not an engineer.
(07:30:05): My background is about a decade working at hedge funds,
(07:30:08): investing,
(07:30:09): and then a decade in fintech startups.
(07:30:11): Most recently,
(07:30:12): I was at Perplexity running the finance vertical and have since been doing a lot of
(07:30:18): consulting and catching up investment firms in this world of AI.
(07:30:23): And so what I wanted to show today is a different way to use Claude that I've been
(07:30:28): really excited about.
(07:30:29): And so a lot of it is Claude is building apps and that's really cool,
(07:30:34): but I found it an incredibly powerful research assistant.
(07:30:39): So I'm gonna share my screen right now,
(07:30:44): and I'm gonna show you how I'm using it in the terminal here.
(07:30:50): So sharing my screen.
(07:30:52): And so what I'm doing is kind of showing you a little bit of like the end product
(07:30:57): is I'm able to create an on the fly dashboard.
(07:31:00): So let's say I'm a financial analyst and I want to produce an earnings preview for a company.
(07:31:07): So Meta is reporting next week and I want to produce this earnings preview.
(07:31:12): With Cloud Code,
(07:31:14): it's really powerful because you can access MCPs,
(07:31:17): you can access all the files on your computer,
(07:31:22): and with one command,
(07:31:24): you can actually create this dashboard.
(07:31:28): What we've done is we've incorporated a lot of tools that keep you out of the
(07:31:34): coding environment by using Cloud plugins.
(07:31:38): What I'm doing is I can just call a plugin using this command here, just the slash command.
(07:31:45): Now I'm calling up a prompt that I already have saved in here.
(07:31:50): and just with a slash command.
(07:31:52): Then I can just say, build a earnings preview for Meta.
(07:31:58): With one click, it's going to create a dashboard basically similar to this.
(07:32:05): The fun thing is it can access MCPs,
(07:32:09): Like I have this Daloopa MCP here, it can access all the files on my computer.
(07:32:15): So it can be my notes, it can be transcripts, it can call APIs.
(07:32:19): So it's really robust.
(07:32:21): And rather than producing an output that is like an old sticky PDF style output or
(07:32:27): an email style output,
(07:32:29): it can actually produce interactive dashboards.
(07:32:33): And so rather than having a static environment,
(07:32:36): I'm producing a dashboard here running in my local host in Streamlit.
(07:32:41): And so it's really interactive.
(07:32:44): So I can move stuff around.
(07:32:46): I can scroll over here and get more detail.
(07:32:49): And this is all coming from an MCP.
(07:32:53): So it's actually really good financial data.
(07:32:56): And it can have tabs.
(07:32:57): This is crazy robust.
(07:32:59): It takes about 20 minutes or so to run this.
(07:33:02): So crazy robust having revenue guidance, beat, miss, track record.
(07:33:08): I can go in here and check out a different tab, guidance analysis.
(07:33:13): It can produce these charts on the fly.
(07:33:15): And the cool thing is that I'll show you in a minute
(07:33:19): how I'm generating these things.
(07:33:21): It's all natural language.
(07:33:23): So I think just say what I want.
(07:33:25): I want a guidance analysis and it will flex the dashboard to be based on the things
(07:33:30): that meta guides to.
(07:33:32): So meta guides like revenue expenses and CapEx and their tax rate.
(07:33:36): Whereas another company may just guide to revenue or it may guide to just EPS.
(07:33:41): It is flexible enough so that you can have a unique dashboard for every single
(07:33:45): company or project that you're working on.
(07:33:49): You can go through here.
(07:33:50): It can even create a mini model in Excel.
(07:33:54): These are specific line items that I care about.
(07:33:56): It can show me where the guidance fits in here, all sourced from an MCP.
(07:34:02): so crazy robust i just want to show you like how i generate all this stuff is is it
(07:34:09): all comes from github and so when you start to use chat gpt and you're starting to
(07:34:18): produce these prompts they get to be like really really long and they start to hit
(07:34:23): like this 8 000 character limit
(07:34:25): And the trouble is you actually want your prompts to be much more robust.
(07:34:29): And so what we can do is we can create a Cloud plugin in a GitHub repo.
(07:34:34): And now I'm saving all my prompts and I'm version controlling them.
(07:34:38): And so you can see I've got this Cloud plugin here.
(07:34:41): I've got skills associated with it.
(07:34:43): So I kind of give it like my investment philosophy,
(07:34:47): what things I care about,
(07:34:48): what data sources it should prioritize.
(07:34:50): And so when I want to use the Dilupa MCP,
(07:34:52): which was all this work here on the left,
(07:34:55): I just said,
(07:34:55): use the Dilupa MCP.
(07:34:57): I'm not writing any code or anything about that.
(07:34:59): When I want to say like, where are the company transcripts?
(07:35:02): I can just go here and tell it specifically.
(07:35:06): So it's now it's asking me like, well, what, what is the directory I should use?
(07:35:09): So I can just go in and give it the exact directory on my computer to go to.
(07:35:13): And so it'll know it'll be in this format.
(07:35:15): So super flexible.
(07:35:16): And then I give it my investing principles.
(07:35:18): Like I like all the goblies.
(07:35:20): I like trajectory.
(07:35:21): I like accelerating businesses, not decelerating businesses.
(07:35:25): So I'm telling you like, what is a great investor?
(07:35:27): Because I find LLMs are just very consensus in the way they look at ideas.
(07:35:32): So I'm turning Claude into this research agent that is running code to produce
(07:35:38): these custom dashboards.
(07:35:40): And so when I want to call these commands, I've got my earnings preview command.
(07:35:44): And so I say like, I want to beat miss track record.
(07:35:47): And so there's my beat miss track record.
(07:35:50): I want to do 12 quarters of beat rate, four quarters of beat rate.
(07:35:55): So now I've got 12 quarters and last four quarters.
(07:35:59): So I just,
(07:36:00): I'm saying the stuff in all this natural language and it's doing all the work is to
(07:36:03): be like,
(07:36:04): well,
(07:36:04): should this be a bar chart?
(07:36:05): Should this be a line chart?
(07:36:08): How should I present this?
(07:36:09): I don't have to talk all about that.
(07:36:10): I can just instruct it how to do this stuff.
(07:36:14): So again, here, guidance analysis, like guidance range versus actual results.
(07:36:18): It chose this graph.
(07:36:19): All I told it to do was compare revenue guidance.
(07:36:24): And I gave it maybe a little hint as to using a line charts.
(07:36:30): And so it was able to do all this work on its own.
(07:36:34): So really powerful combination of cloud code,
(07:36:36): GitHub,
(07:36:38): and streamlet apps to produce crazy analysis.
(07:36:42): So I'll pause that and see if there's any questions.
(07:36:44): Yeah.
(07:36:44): Yeah.
(07:36:45): I was going to say,
(07:36:45): you know,
(07:36:45): Brooker,
(07:36:46): you grew up,
(07:36:47): if you will,
(07:36:48): at like some of the world's greatest hedge funds and have had like a career in
(07:36:51): finance before you were at Perplex feeling the finance vertical.
(07:36:56): How long would this have taken you to do as a hedge fund analyst in the earlier
(07:37:01): side of your career?
(07:37:03): my god it's it's crazy to imagine because this work for an earnings preview is like
(07:37:09): five hours of work compiling all these different sources together into one
(07:37:13): dashboard and my dashboard would be like a word document or a powerpoint
(07:37:18): presentation it wouldn't be interactive it wouldn't be live and so now
(07:37:24): i can do all this work and i can have a unique dashboard it's like you're building
(07:37:28): a a bloomberg or like a dashboard system like a koi fin uh for your entire process
(07:37:36): um and so now like it's the software development is combined with um the query and
(07:37:42): and the actual output which is really cool i mean it strikes me that this is uh so
(07:37:48): clearly the future of how we might think of engaging with the financial markets
(07:37:53): and the public market specifically, what are you most excited to be working on this year?
(07:37:59): And can you tell us a little bit about the,
(07:38:01): you have a platform where you have a strategy that you run.
(07:38:05): Can you tell us a little bit about that?
(07:38:07): Yeah, so I can share.
(07:38:09): The reason I'm so excited about this,
(07:38:11): and I'll share my screen again,
(07:38:12): what gets me going on this cloud code is if you can see my screen,
(07:38:19): the issue is when you use ChatGPT or the web-based interfaces,
(07:38:25): you're pretty much only getting 20 to 30 minutes of compute.
(07:38:29): Ever since the start of 2025,
(07:38:32): we've expanded to offer LLMs significantly more time to process and create answers.
(07:38:40): But the web apps have kind of stuck at 20 minutes.
(07:38:43): And so there's this huge gap.
(07:38:45): And I think that's why ChatGPT is by cloud code is getting so much attention right
(07:38:50): now is because this gap is so massive.
(07:38:54): And so you're just scratching the service on compute.
(07:38:56): And so I'm really excited about like this line going up.
(07:39:00): And so you can create so much more analysis and let it run for so much longer on your machine.
(07:39:07): And so, yeah, I do.
(07:39:10): I do investing and I actually publish all my trades and all my ideas and all my
(07:39:16): research on X.
(07:39:18): And so you can actually see all my trades.
(07:39:20): It's totally transparent.
(07:39:21): And so I'm really excited with that with Autopilot.
(07:39:24): It's the same people who created the Pelosi tracker.
(07:39:26): And so, yeah, I love talking about these ideas and sharing the analysis there.
(07:39:31): We had a question in the comments.
(07:39:34): Where do you get your fundamental data from?
(07:39:38): There's lots of good places.
(07:39:40): If you're like an institutional investor, DeLupa is really good.
(07:39:44): On the retail side, I'm partial to Perplexity Finance.
(07:39:48): You can grab transcripts there really easily.
(07:39:51): On Perplexity Finance, you can actually, I'll share my screen really quick.
(07:39:56): on perplexity finance, you can actually download all the financials as well.
(07:40:00): So I'll just look here.
(07:40:03): So if I like my favorite company of mine is ICE.
(07:40:05): They are a monopoly of exchanges.
(07:40:08): So you can go here through earnings and you can download transcripts.
(07:40:12): And then you can go through financials and you can download all the financials.
(07:40:16): And so all this stuff, which used to be gated, you can access all this information.
(07:40:21): You can even get research reports here as well.
(07:40:25): So crazy powerful.
(07:40:26): If you want to do it in API access, you can do like a fiscal AI is really good.
(07:40:32): And you can build a skill inside cloud code that will be able to allow LLMs to access APIs.
(07:40:41): So it's really easy to start layering and all these different data sets once you
(07:40:45): have this base set up and running.
(07:40:49): Um,
(07:40:49): another question here is,
(07:40:51): uh,
(07:40:51): you know,
(07:40:52): this is sort of maybe going off of your perplexity resource.
(07:40:55): Any other favorite places to either learn anything financial industry related or
(07:41:01): vibe coding favorites?
(07:41:04): I've been paying a lot of attention to what Claude is releasing these days and what
(07:41:08): MCPs are coming up next.
(07:41:10): Um, but the X is the best source for the financial news.
(07:41:16): And I would do a little plug for anyone that is particularly interested in this
(07:41:21): topic and maybe in the public markets in particular,
(07:41:24): we will be doing more content courses and workshops around this.
(07:41:30): So stay tuned and you'll see more of Brooker around that topic.
(07:41:36): Let's see if there are any other questions.
(07:41:39): Brooker, anything else you would like to share?
(07:41:41): No, that's it.
(07:41:42): I was just really excited to share that with you guys.
(07:41:44): I think there's a lot of stuff you can add because I think it's really important to
(07:41:48): start building this IP of these prompts.
(07:41:50): I think they're so valuable and GitHub is a great place to store that.
(07:41:54): Yeah.
(07:41:55): Thank you.
(07:41:56): Thank you.
(07:41:56): So excited to be doing more of this together.
(07:41:59): And again,
(07:42:00): if you're really into this particular side of the world,
(07:42:04): stay tuned,
(07:42:05): follow,
(07:42:05): and there will be much more coming.
(07:42:08): Back to you, Austin.