Claude Code for Investment Research (Part 4): The Obsidian Layer That Makes Everything Compound
How to make Obsidian the brain that holds every article, every concall, and every market thread you've ever read. The layer that makes Claude Code compound instead of forget.
Ninety days ago, I opened a terminal for the first time and typed claude. Today, the same stack runs my entire investment research workflow. Screening, scraping, modeling, drafting, monitoring. The terminal stopped being intimidating somewhere around day three.
Three newsletters in, you’ve seen most of the moving parts. Part 1 was the confession, why a finance guy would even open a terminal in the first place. Part 2 was the full setup: Claude Code, Cursor, Obsidian, Playwright, Firecrawl. Part 3 was the first real workflow, decoding insider trades.
If you’re new here, those three are worth reading before this one. They build the system. This newsletter is about the part of that system most readers underestimate, and the part doing the most quiet work in the background.
Obsidian.
1. The Bottleneck Was Never the AI
Picture a Monday morning. You’re scrolling X with your coffee. Someone shares a research piece on the semiconductor capex cycle. You read it. It’s good. You make a mental note that this’ll matter for the chip names on your watchlist.
Friday rolls around. You’re researching one of those names. You remember the article. You don’t remember where you saw it. You scroll X for ten minutes. You search your bookmarks. You give up. You do the analysis without it.
Multiply that by six months of reading. That’s the real bottleneck in AI-powered research. Not the AI. The collection.
Every analyst I know has tried to build a personal research library. Most have failed. The reason isn’t laziness. It’s that the act of filing is exhausting. Bookmark the page. Open a folder. Rename the file. Pick a tag. Upload it somewhere. Six months later, find it again. The first time you do it, it takes ninety seconds. The fifth time you do it that morning, you stop doing it.
The AI tools we’ve been talking about in this series, Claude Code, NotebookLM, the rest of the stack, all assume you’ve solved this problem. They assume the right document is already in the folder when you ask the question. They don’t help you put it there.
That’s the gap.
Now picture the same Monday. You read the same article. You hit one button in your browser. The article is already a markdown file in your research folder. You didn’t pick a name. You didn’t pick a tag. You didn’t decide where it goes. Friday comes. You ask Claude Code: what have I read on this sector recently? It pulls the article. It pulls the concall transcript you clipped three weeks ago. It pulls the chart you saved last month.
Now zoom out. Six months. A year. Two years.
Every article you’ve ever read. Every concall. Every chart. Every newsletter. Every X thread. All in one folder, all in markdown, all readable by Claude Code in a single query.
Your research stops disappearing. It starts compounding.
That’s what Obsidian does. The next sections are about how.
2. Obsidian, In One Paragraph
Obsidian is a free app. You download it. It points at a folder on your computer. Every note you create is a plain text file inside that folder, written in markdown. The folder is the product.
That sounds boring until you realize what it means. Markdown is the format AI models read most easily. Plain text on disk is the format Claude Code can write to, edit, and search without any bridge or API. A folder of markdown files is, by default, the most AI-friendly research library you can build. Obsidian’s only job is to make that folder pleasant to live in: a sidebar, a search bar, a graph view of how your notes connect, a plugin system for the things the core app doesn’t do.
If you read Part 2 of this series, you’ve already installed it. You created a vault with three folders, raw/, wiki/, and output/, plus a CLAUDE.md rulebook that tells Claude Code how to behave inside it. If you haven’t done that setup yet, do it now, 10 minutes of work, and the rest of this newsletter assumes it’s in place.
Prefer to watch? I recorded a forty-three-minute walkthrough of the entire system, including live use cases and the parts that are hard to show in writing. The newsletter stands alone, but the video shows what the screen looks like.
What I want to spend the rest of this piece on is what happens after the install. Because Obsidian by itself is just a note-taking app. Obsidian wired into Claude Code, with one browser extension and one design principle, is the part doing the quiet work in the background.
The browser extension is called Web Clipper. The design principle is that you don’t decide where anything goes. You’ll see what I mean.
3. What "Any Data" Actually Means
The Web Clipper turns any page in your browser into a markdown file in your vault. That’s the whole feature. But “any page” is doing a lot of work in that sentence. Let me show you what it actually covers.
Articles and prose. This is the obvious one. You read a Bloomberg piece, an FT analysis, an X thread, a Substack newsletter, a Seeking Alpha take. One click and it’s a markdown file in raw/ folder. The Web Clipper strips the navigation, ads, and sidebar junk. What lands in your vault is the article body, the title, the author, the source URL, and the date, clean enough that Claude Code can read it without parsing through a wall of cookie banners. If the page had images, install the Local Images Plus plugin once, and every image gets downloaded into the vault alongside the text. Your clippings stay intact even if the original page goes down six months from now.
Structured tables. This is the one most people don’t expect to work. Screener.in’s financials page. NSE’s top 500. Finviz’s insider trade dashboard. The S&P 500 sector grid. An ETF list from any exchange. Anything that renders as an HTML table on a web page becomes a clean markdown table in your vault. I clipped the entire NSE ETF list one Saturday afternoon while I was thinking about where to deploy some capital. Two clicks, twenty seconds, and Claude Code could read the whole universe.
Recurring feeds. This is the one that compounds the fastest. You subscribe to a Bloomberg weekend email. You get an insider trade dashboard delivered weekly. You follow ten X accounts that post good market threads. Every one of those is one click away from being a markdown file in your vault. After a month, you have a structured archive of your information diet. After a year, you have something no aggregator can sell you, your own curated stream, organized the way you read.
The principle behind all three is the same. The thing you’re already doing when you read on the internet is the act of curation. You decide what’s worth your time. You read it. That decision is the work. The Web Clipper makes sure the work doesn’t evaporate the moment you close the tab.
4. The Quiet Discipline
Most knowledge management systems fail in the same way. You install the app, you set up the folders, you clip a hundred things in the first week, and then six months later you can’t find a thing because you stopped organizing the moment the act of organizing got boring.
This system doesn’t fail that way. Not because of better discipline. Because the discipline is delegated.
The CLAUDE.md rulebook you set up in Part 2 is the part doing the work. It’s a plain text file at the root of your vault that tells Claude Code exactly what its job is. Claude Code is the librarian. The raw/ folder is its inbox. The wiki/ folder is its domain, where it writes summaries, builds topic indexes, and links related ideas across articles. The output/ folder is where finished work lands. You don’t file anything. You don’t tag anything. You don’t decide what goes where. You clip. Claude Code files.
Once a week or whenever you remember, you open Claude Code and type one word. Compile. It reads every file in raw/ that hasn’t been processed yet, decides which topic each one belongs to, writes a structured summary, files it in the right wiki folder, and updates the master index. By the time you finish your coffee, your inbox is empty and your library is bigger.
That’s the whole loop. Clip. Compile. Query. Audit. Four verbs, covered in detail in Part 2. The thing I want you to notice is what those four verbs don’t include. They don’t include “organize.” They don’t include “tag.” They don’t include “remember where you put it.” Those jobs belong to Claude Code now.
Here’s the compounding part. Every time you ask Claude Code a research question, you can tell it to save the answer back into the wiki as a new note, linked to the sources it used. Based on everything in the vault, what’s the bull case on Indian specialty chemicals? Save your answer as a new wiki article and link it to the sources you cited. Now the question becomes a note. The note becomes a source for the next question. The library doesn’t just grow, it grows in the direction of the work you’re actually doing.
After ninety days of running this, I can ask Claude Code what have I learned about AI infrastructure capex this quarter? and get a real answer in under a minute, drawn from articles I’d half-forgotten clipping. The system remembered for me.
That’s the quiet discipline. You don’t have to be a better filer. You have to be a clipper. The rest takes care of itself.
5. A Tuesday Afternoon, Three Minutes, Seventeen Companies
Here’s what the system does on an ordinary day.
A few weeks ago I clipped an article I’d written myself, called I Warned You About Infosys. Here’s How to Find the Next Winners. The thesis was simple. Indian IT services were structurally exposed. AI agents had collapsed the cost of building software to near zero. The bottleneck had shifted from building the product to growing, hosting, and securing it. Every new vibe-coded app that launched would have to pay what I called a growth tax, to advertise itself, to host itself, to secure itself, to retain its users. The real winners weren’t the chipmakers or the model labs. The real winners were the tollbooth operators downstream of every new app.
The article had a full institutional research prompt at the bottom. The kind of prompt you’d run inside ChatGPT or Gemini Deep Research to map the public companies that benefit from each layer of the growth tax. I’d written that prompt for readers to copy and use. I’d never run it on my own stack.
Last Tuesday, I did.
I opened Claude Code. I typed one line. Find the article I clipped about Indian IT and the growth tax thesis, then run the institutional research prompt at the bottom of it. Save the output in output/.
Three minutes later, Claude Code handed me a memo titled The Infinite App Supply Shock: Mapping the “Growth Tax” Tollbooths. Executive thesis. Four modules. A ranked watchlist of seventeen public companies, Meta and Alphabet for the Attention Tax, Cloudflare and Amazon for the Plumbing Tax, Okta and Palo Alto for the Trust Tax, Twilio and HubSpot for the Habit Tax. Each name with the direct causal link to the thesis, three KPIs to track in their next 10-Q, a pricing model assessment, and a private market threat layer flagging the unlisted competitors (Stripe, Vercel, OpenAI) attacking each tollbooth from below.
It also wrote an extinction list. Five public companies whose margins would compress hardest as the thesis played out. Salesforce. Accenture. The legacy IT services names. With the bear thesis on each and the conditions that would invalidate it.
This is not a recommendation. It’s a demonstration. The point isn’t the seventeen names. The point is the workflow.
I didn’t have to find the article. I didn’t have to copy the prompt. I didn’t have to paste anything into a chat window. I didn’t have to specify which model to use, which API to call, or where to save the result. Claude Code went into the vault, found the article by topic, read the embedded prompt, executed it, and filed the output. Six weeks earlier, I had read that article on the train. Now Future Me, in three minutes, had operationalized it.
That’s the loop. Read on Monday. Forget by Friday. Operationalize months later, in three minutes, on a question you didn’t even know you’d ask when you clipped the piece.
The article is in the vault. Claude Code can find it. That’s all the system needs.
6. The Same System, The Other Direction
Yesterday morning, the most-watched 13F filing of this quarter dropped.
Leopold Aschenbrenner’s Situational Awareness LP, the AI-thesis fund that has become one of the most-followed books on Wall Street, filed its Q1 2026 disclosure with the SEC. The financial press lit up within hours. Yahoo, Benzinga, Bankless, the Motley Fool. The framing was sharp and the framing was uniform: a $8.5 billion bet against the chip complex.
Section 5 was about a clip from six weeks ago. This section is about a filing from 36 hours ago. Same system, opposite direction.
I opened Claude Code last night. I typed six lines.
It’s 13F season. Aschenbrenner’s Situational Awareness fund just filed yesterday and the market is talking about his chip shorts. Pull his latest 13F. Also pull Ackman, Einhorn, and Pabrai for context. Save each one in
raw/13f/as a clean markdown table. Then inoutput/, write me a one-page memo on what Aschenbrenner is doing and how it compares to what the value guys are holding. What’s he betting against? What’s he betting on? What changed since last quarter?
That’s it. Six lines. No URLs. No source specification. No memo format brief. No deliverable schema.
I didn’t tell Claude Code to use Firecrawl. I didn’t tell it where to find 13F data. I didn’t pick between Dataroma, WhaleWisdom, 13f.info, or SEC EDGAR. I didn’t ask for a specific output format. I didn’t even tell it which models to compare against, the value guys was enough.
Two minutes and ten seconds later, the vault had four new markdown files. The Aschenbrenner filing. Bill Ackman’s Pershing Square. David Einhorn’s Greenlight Capital. Mohnish Pabrai’s three-position energy book. Each one parsed into a clean structured table. And in output/, a one-page comparative memo titled Aschenbrenner vs. The Value Guys Q1 2026.
The memo itself isn’t the point of this section. I’m not going to walk through its conclusions here, what Aschenbrenner is actually betting on, whether the chip puts are bearish or hedging, how Ackman’s Microsoft rotation compares to the others, that’s a research piece in its own right. I’ll come back to that in a future newsletter, once I’ve done the cross-check work that any honest AI-assisted analysis demands.
What I want you to notice here is the workflow.
Six lines of plain English. Claude Code picked the sources. Claude Code picked the tools (Firecrawl, in this case). Claude Code decided what a “clean markdown table” should look like for a 13F. Claude Code decided how to structure a comparative memo without ever being shown one. Two minutes and ten seconds from prompt to four parsed filings and a draft analyst memo on my disk.
The same vault that compounded six weeks of inbound reading in Section 5 also caught up to a sixteen-hour-old market event in Section 6. Web Clipper handles the inbound flow. Firecrawl and Playwright handle the outbound flow. Same folder structure. Same rulebook. Same query layer.
Two cognitive modes. One library.
7. Where This Leaves the Stack
Ninety days into running this stack daily, here’s how I think about each piece now.
Claude Code is the operator. It’s the thing you talk to. It reads, writes, scrapes, files, drafts, and refuses to do work it can’t ground in something on your disk. Everything else exists to make Claude Code’s job easier.
NotebookLM is the citation-grade focus tool. Tightly scoped, deeply sourced, perfect when you’re going deep on one company or one sector and you want every answer to come with a footnote.
Obsidian is the always-on research brain. It’s not for one company or one sector. It’s for everything you read, everything you clip, everything you’d otherwise forget. It’s the place Claude Code looks first when you ask a question that touches your past reading. And it’s the place every Firecrawl scrape and Playwright run lands when you ask a question that needs the live web.
Cursor is the workspace. The terminal where Claude Code lives, the editor where your CLAUDE.md rulebook lives, the place you actually sit down and do the work.
Four tools. One library. The library is the part most readers underestimate.
Part 5 of this series will go deeper on the outbound flow, the Playwright side of the stack. How to point Claude Code at any web source on a schedule, how to wire portfolio monitoring into your vault so the system surfaces changes before you ask, and how to build a personal research feed that updates itself.
For tonight, the system is enough.
One announcement before I close.
I’ve launched the Alpha with AI Club, a space for analysts, investors, and founders who are serious about learning AI for investment research. Not casual. We have members from 25 countries already, sharing workflows, comparing setups, and working through the kind of problems this newsletter walks through.
The free tier lets you join the community and access the open discussions. The premium tier gets you our internal playbooks, direct access to me for questions, and the workflows I might not publish in the newsletter.
If you’ve been following this series and want to go deeper than what fits in a Substack post, join the club here.
Only for people who are serious about this work. If that’s you, the door is open.
🤝 Help Us Grow This Circle
Thank you for reading and supporting Alpha with AI. If you share this edition with even one person who might find it valuable, it means the world to us and helps this project reach those who need it most.
At Shikshan Nivesh, our goal is simple, to make financial research faster, smarter, and more accessible.
We believe investing should start with understanding and every newsletter we write is built to reflect that.
Written by Shubham Borkar | Research & Insights by Shikshan Nivesh AI Team
Financial Clarity. Insightful Ideas.
Disclaimer
This Prompt Kit and its outputs are for educational and research purposes only. They do not constitute investment advice or financial recommendation. Always verify disclosures and consult qualified professionals before making investment or business decisions.