- Fintech Brainfood
- Posts
- AI Agents will be Wall St Firms
AI Agents will be Wall St Firms
Meet the hedge fund run entirely by AI Agents. Plus, Nubank does Stablecoin rewards, and Coinbase does Bitcoin-secured Stablecoin lending.
Welcome to Fintech Brainfood, the weekly deep dive into Fintech news, events, and analysis. You can subscribe by hitting the button below, and you can get in touch by hitting reply to the email (or subscribing then replying)
Hey Fintech Nerds đź‘‹
What happens when the world’s most known digital-only bank adds Stablecoin rewards? They become the new normal. And every other bank CEO scrambles to figure out how to copy. (More in Things to Know 👀).
Fintech had a HUGE week. Coinbase does Bitcoin-backed loans, Wise crushed again, eToro will IPO, Freetrade got acquired, Block got fined, and nobody wants to partner on Apple Card. Sign of the times.
The $TRUMP meme coin is as fascinating as it is gross. Longer take from me here. A memecoin rant is coming in a few week’s time.
Meanwhile, Goldman says AI can write 95% of an IPO prospectus in 5 minutes.
AI is eating Capital Markets, just not in the way you expect. (Your Rant this week)
PS. For those coming to Utah next week, I’ll be in town on the 23rd and 24th for the Fintech Exchange at the Stenna Center. See you there?
PPS. Did you know I have a podcast about Stablecoins? You should probably check it out if the subject has caught your attention.
Here's this week's Brainfood in summary
đź“Ł Rant: AI Agents will become Wall St Firms
đź’¸ 4 Fintech Companies:
đź‘€ Things to Know:
đź“š Good Read: Lynn Alden December Newsletter
If your email client clips some of this newsletter click below to see the rest
Weekly Rant đź“Ł
AI Agents will be Wall St Firms
On the face of it this tweet is utterly mind-blowing.
I’ve been building a real-world AI hedge fund.
It's open source so you can learn + build too.
The hedge fund has 6 agents:
1 • market data agent
2 • quant agent
3 • fundamentals agent
4 • sentiment agent
5 • risk manager agent
6 • portoflio manager agentThe hedge fund… x.com/i/web/status/1…
— virat (@virattt)
4:54 PM • Dec 21, 2024
Someone build a hedge fund with AI Agents.
Well, not quite.
Look closer; you'll see it's doing pretty basic things with zero competitive edge. Most AI experiences suck. Most test agents do. But threads like this are catnip for VCs, bankers, and founders alike because it has never been easier to build an MVP that looks like a real financial institution.
Here's the thing - our pattern recognition has shown countless times that what starts as a basic product can quickly become an industry behemoth.
The idea of an AI-first, or AI-only hedge fund is not science fiction. o3, and perhaps o4 will move us to a phase where AI consistently out-performs humans at human tasks (if the leaks are to be believed).
When that happens the competitive edge in capital markets shifts from HFT and quant, to analysis.
Maybe the next Goldman Sachs won't be built by bankers but by AI-native engineers who understand how to make agents actually perform in production.
Today we cover
Thesis: The Three Paths
Why capital markets are ideal for “Agentic” work
AI Agents are well-packaged LLMs for a complex domain
Capital Markets are the perfect domain for AI
And there are early examples its working
AI Changes the moats for capital markets firms and tips the scales to innovators
The wild future can be seen by looking at Crypto x AI
Thesis: Three Paths for Capital Markets and AI
There are three paths in front of us
The AI-Augmented Firm (inevitable)
The AI-First Firm (likely)
The AI-Only Firm (wild but not impossible)

The three paths to disruption of Wall St
The AI-Augmented firm is a productivity gain.
What: Analysis is simplified by Generative AI summarization, and generated outputs
Competitive advantage: Data-moat and High Frequency Trading (HFT) gives a statistical competitive advantage
Talent Mix: Instead of armies of junior analysts, you'll have AI agents doing first-pass analysis
Clients: Traditional prime relationships
The AI-First firm is a competitive advantage.
What: New analysis is discovered by Generative AI
Competitive advantage: Net new AI-driven analysis becomes the competitive advantage
Talent mix: Former analysts build new firms with tools they couldn’t access in their day job
Clients: New prime or hybrid prime relationships
The AI-Only firm could be a real revolution.
What: Analysis is performed by, and trades are executed 100% by AI over multi-year and long term time horizons
Competitive advantage: Combines HFT like speed with net-new analysis to perform “AlphaGo” like trades that humans wouldn’t see over the long term. Operating at machine speed and 24/7
Talent mix: 100% AI?
Clients: Direct access for market participants (and retail if on Crypto rails)
Perhaps the next Goldman Sachs won't be built by bankers adopting AI - it'll be built by AI agents that grew into banks.
AI Agents are LLMs-as-Product for a problem domain.
Everyone's got a definition of AI agents. Most read like they were written by a committee of product marketers:
"AI Agents are software that can operate autonomously, pursuing complex goals by understanding instructions, setting tasks, and adapting to conditions."
For me, an AI Agent is just a well-packaged LLM-as-product that can do useful things.
Turns out that’s a 100x harder than it sounds.
I’ve used this example before. Imagine you ask your AI agent to find a flight to Antarctica and book the cheapest one. It might look like this:
Now get access to all of that confidential data and have it work consistently.
An AI agent is a system that can take an input (like "analyze this company"), break it into steps ("pull the financials, check the competition, review the management team"), and execute those steps without constant human hand-holding.
Think of agents as an LLM that calls a bunch of APIs within set guardrails. The LLM is the brain that figures out which APIs to call and in what order.
Building a demo that can follow a script is easy. Building something that can handle edge cases, maintain consistency, and actually deliver value? That's where things get interesting.
Capital markets are the perfect domain for productized AI they have
Lots of secure, confidential data sets (e.g., data rooms)
Very large, impossible to fully review data sets (e.g. supply chain, news, macro)
A very complex problem domain (e.g,. The bond market)
A high cost of failure if the LLM makes a mistake
Capital Markets are the Perfect Case Study for Agentic AI
While Fintech disrupted simpler products, capital markets remained untouched - until now.
Despite the wealth involved, most contracts are agreed upon over email, PDFs dominate deal flow, and Excel is still king.
While Fintech disrupted simpler financial products, capital markets have been largely untouched by the Fintech revolution until now. This high-margin corner of the industry is filled with jargon, specialism, and a relationship-driven culture that's ripe for change.
What makes capital markets analysts perfect for AI replacement?
They wade through massive unstructured data sets (PDFs, emails, spreadsheets)
They navigate complex jargon and industry-specific knowledge
They need to make sense of market data scattered across websites
They spend countless hours on data entry and summarization
Analysts are typically highly-paid, rare skill sets that cannot be outsourced
Generative AI is great at these sorts of problem spaces.
Early Signs AI is Working in Capital Markets
GenAI is already eating into traditional analyst workflows:
I'm not sure people understand how important this is:
Goldman Sachs CEO says that AI can draft 95% of an IPO prospectus “in minutes.”
White collar jobs will go first.... and fast.
— John LeFevre (@JohnLeFevre)
12:37 AM • Jan 17, 2025
1. Equity Analysis. AI can scan through earnings calls, SEC filings, and market data. It can generate preliminary analysis and flag key points for human review. Example companies like
Finster, Quill or Alphawatch update any financial model with market data in real-time.
2. Credit Workflows. AI agents can process loan applications and documentation. They can flag risk factors and anomalies and generate standardized credit memos. Companies like Auquan or 8vdx can build credit memo’s, and review a portfolio for KPIs against their contracts or can extract data into excel models.
3. Private Equity. AI Agents can trawl through 1,000s of data rooms and surface possible due diligence risks, find buyers or companies looking to sell, and cross reference this with internal data. E.g. Keye. and Decisional.
Every Capital Markets workflow has some AI company re-packaging LLMs to attack it.
AI Changes the Moats for Capital Markets firms
What will be the moat for the firms that can enter capital markets, either as a new hedge fund, private credit or asset manager?
​​Today's moats are crumbling:
Quant talent becomes less differentiating when models like o3 are starting to outperform humans
Scale and brand matter less when AI can do the work of 100 analysts
Statistical advances are an arms race with well-funded competitors making marginal gains vs each other
But new moats are emerging:
For AI-Augmented (Traditional) Firms:
Integration with legacy systems and proprietary data sets gives them an insight advantage
Existing customer relationships give them economies of scale and certainty over the ability to invest
Regulatory compliance frameworks are well-known
For AI-First (New) Firms:
Speed of deployment of new ideas, models, and tools gives them the ability to price or discover deals that others might miss
Clean-slate architecture gives them the ability to try new things and attract frustrated analyst talent from elsewhere
More efficient cost structures and lower overheads potentially let them offer a pricing differential to disrupt the market
For AI-Native Firms (Crypto):
The ability to operate autonomously means they don’t need to sleep and won’t fall for human biases or failures
Direct market access means no gaps between identifying an opportunity, trading, and settlement
Over time their market data would drive self-improving performance that’s harder/slower for humans.
Look, I’m not saying Crypto AI Agents are likely to take over the world.
But I am saying:
The Most Interesting AI Agent Experiments Are in Crypto
Very few regulated firms are going to experiment with letting their AI Agents trade significant sums of money. But that’s live today in Crypto.
AI Agents x Crypto are this cycle's new hype.
And the token growth mirrors DeFi in 2021.
Wherever there’s hype, there are good reasons to be skeptical. There will be scams, rug pulls, and hacks. Buyer beware.
However, from the soup of seemingly gross Crypto, weirdness often crawls market-changing innovations. Consider that:
Stablecoins are the biggest Fintech topic.
We have a Bitcoin and Ethereum ETF.
Polymarket was a fringe idea, and now it’s a case study.
To bring this to life, consider ai16z. An on-chain treasury and wallet led by AI-Agents who make trades (and some who even write their thesis and insights on X).
This DAO and fund has
An AI agent that makes all trading decisions
Has $10M+ assets under management
Token up 1,660% since November
The GitHub repo for the AI framework that ai16z is built on "Eliza" just hit #1 globally with 7,100 stars in weeks. These agents have an LLM for a brain, but also direct access to on-chain data and the ability to respond to and trade market conditions.
Don’t look at the coin price today, or next year, look at the pattern.
Ask yourself
What happens when AI agents start running real money and building wealth?
When they become the LPs in funds?
When they're making allocation decisions?
The big players won't notice at first. They'll be too busy with quarterly earnings and committee meetings. But underneath, a new kind of financial institution is being born - one that's AI-native from day one.
The AI-first firm.
The Real Revolution
Ignore the “This is Wild! Google just killed Capital Markets with their new AI”
Start thinking about what happens when the AI-first asset manager passes $1trn of AUM.
Or if an AI-Agent wallet passes $100m AUM?
Building something like that in a way that doesn’t blow up?
That’s a lot harder than writing a viral Twitter thread.
ST.
4 Fintech Companies đź’¸
1. Kolena - Capital Market Analyst AI
AI Agents for finance, with 100s of use cases, like private equity diligence, analyzing audit evidence, earnings calls, and cross-referencing regulations. Kolena claims to be "the only AI Agent that doesn't fail at Maths tasks"
🧠What's the moat if everyone is doing this? Kolena's IP is "restructured," where it takes unstructured data and categorizes it first. I imagine this is how they identify maths, pull it away from LLMs, and move towards traditional software approaches. Bit by bit, AI is eating the analyst labor-intensive tasks of capital markets.
2. Synthera - Synthetic capital markets data.
Synthentic financial markets data so professional investors can test their portfolios on thousands of unseen market scenarios and unlock novel insights previously beyond reach. Synthera can fill gaps in historical market data and identify non-linear correlations to simulate risk and reward for investors.
🧠Low-key is super useful, but forecasting is hard without data for scenario planning at scale. Your forecast is as good (or as limited) as your data set. It's hard to model what hasn't happened yet. If you can build enough synthetic data, you can model that out. This is the other half of the story about AI eating capital markets. A lot of what teams did in-house is becoming unbundled.
3. Briefcase - AI - Month End Automation for Finance teams
Briefcase automatically categorizes suppliers, calculates tax returns, and infers supplier rules from user behavior. It helps calculate deferred revenue and pre-payments and separates duplicate invoices vs receipts.
🧠Accounting AI is so hot right now. I haven't seen one focused on the UK market yet, and I cannot wait for this. The UK solopreneur business space is a giant mess. The UK doesn't really have tax auditors like H&R Block that help with the lift; it's all accountants. The sooner this is baked into everything the better.
4. Kast - The USD Stablecoin Neobank for APAC
Kast helps consumers in hundreds of countries and is focused on APAC markets with large populations, such as Vietnam, Thailand, and the Philippines. Users can save, spend, or receive stablecoins, and the service works with Google Pay and Apple Pay. The digital card costs $18 monthly, and the physical credit card costs $888 annually.
🧠The US Dollar has the greatest product market fit of any product in history. When I speak to my friends in the global south, Stablecoins are a much bigger deal than when dealing with folks who only live and transact in the US. A Stable currency and the ability to hold and spend dollars is huge. Local regulators have tried to push back against this with limited success. I expect a flip where more markets try to collect taxes and formalize this product type.
Things to know đź‘€
Nubank will reward holders of the digital dollar, USDC, with a 4% yield on all balances above $10 in Nubank Cripto wallets. Nubank users can now buy, sell, hold, swap, or earn yield across assets like BTC, ETH, SOL, UNI, and USDC.
🧠What happens when the worlds most well known digital bank offers Stablecoins? They become the new normal. I guarantee several large bank CEO’s saw this story and will be asking their teams what to do next.
🧠Stablecoins have product-market-fit outside the USA. Stability is sexy if you live in a market with high inflation and currency volatility. And guess what, Nubank has a lot of customers in markets where that’s true.
🧠There’s a really cool alert feature if the Brazilian Real moves in high % against the US Dollar. Imagine you’re a user and you see that the Brazilian Real just jumped 1% against the dollar in a day, you might want to move more of your balance into USDC or vice versa.
🧠Nubank is standing on the shoulders of great work by the Brazilian central bank. The Banco Central do Brasil does not get the credit it deserves for doing a lot of hard yards on policy and regulation to make this possible. The staff are young and innovative and communicate with companies to help develop new policies in a less adversarial atmosphere than you’d be used to in the USA.
Coinbase users can borrow the Stablecoin USDC at “competitive rates” with flexible, open-ended terms.
The loans are powered by Morpho, an open-source lending protocol on base (the Ethereum L2). Users can borrow up to $100k instantly, secured by their Bitcoin.
There are no fees, and the rate varies in real-time. When you take out a loan your Bitcoin is tokenized as wrapped Bitcoin and passed to Morpho as collateral for your loan.
🧠You can’t afford to ignore DeFi. Nubank offers Stablecoin rewards, Coinbase offers DeFi loans. What is your company doing?
🧠DeFi loans will become available to Coinbase 75M+ users. Coinbase has invested heavily in building a secure, fast DeFi layer above Ethereum called base. On this, teams have built DeFi projects that make modern digital finance products that exist without a bank
🧠This is how the ultra-wealthy live. If you’re worth a few billion you never sell assets, you borrow against them for cash. This is now available to anyone who did well in Crypto, which is a lot of people who held through the last crash.
🧠What if this is the future of private credit? DeFi lending looks like private credit but for the mainstream and is available to anyone with compatible software. Will private credit funds get involved now there’s a new administration?
🧠Private credit hasn’t gone global but this could unlock it. In most of the world Private Credit didn’t grow exponentially post-2008 as it has in the US (due to regulations, culture etc). The US private credit funds could have a field day if this model can be more confidently priced scaled with liquidity.
Good Reads đź“š
This Macro readout is so on point. My takeaway is that the US stock market is currently trading at 207% of the US GDP, which is an all-time record high. Stablecoins are emerging as a way for businesses in the global south to hold on to purchasing power, and this will only accelerate.
Tweets of the week đź•Š
FICO DRIFT - you now need a 802 score for a consumer vehicle loan at Bank of America, at least based on the averages
— Sonali Basak (@sonalibasak)
2:23 PM • Jan 16, 2025
How to compare your eng team's velocity to industry benchmarks (and increase your velocity):
Step 1: Send your eng team this 4-question survey to get a baseline on key metrics:
docs.google.com/spreadsheets/d…You can use any surveying tool to do this—Google Forms, Microsoft Forms,… x.com/i/web/status/1…
— Lenny Rachitsky (@lennysan)
7:55 PM • Jan 15, 2025
That's all, folks. đź‘‹
Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)
(1) All content and views expressed here are the authors' personal opinions and do not reflect the views of any of their employers or employees.
(2) All companies or assets mentioned by the author in which the author has a personal and/or financial interest are denoted with a *. None of the above constitutes investment advice, and you should seek independent advice before making any investment decisions.
(3) Any companies mentioned are top of mind and used for illustrative purposes only.
(4) A team of researchers has not rigorously fact-checked this. Please don't take it as gospel—strong opinions weakly held
(5) Citations may be missing, and I've done my best to cite, but I will always aim to update and correct the live version where possible. If I cited you and got the referencing wrong, please reach out