Carrington Labs, Zest AI, & Coinbase.
Happy Monday, Fintech Takers!
I hope you had an enjoyable and productive weekend.
Mine was bursting at the seams with both fun and productivity.
One highlight — gathering up all the leaves, which is a chore that I normally despise, but has become much more enjoyable now that my dad has a high-powered electric leaf blower, which he graciously allows me to borrow.
Leaf blowing is way more fun, though perhaps less efficient than raking, which is kinda strange if you think about it.
Is it possible that the midlife crises that people go through are fundamentally just about swapping out something efficient (the minivan) for something less efficient but more fun (the sports car)?
I’m not much of a car guy (and I love my minivan FWIW), but perhaps a surge in power tool investments is in my future?
TBD …
Anyway, it’s Builders Summit week, so I will be in the mountains (and away from the internet) for much of this week. But I promise to provide a summary of all the amazing content and conversations in Friday’s newsletter.
- Alex
Sous-Bois (Le Huelgoate) (1892) by Paul Serusier.
3 FINTECH NEWS STORIES
#1: Carrington Labs Copy anchor linkCopied
What happened?Copy anchor linkCopied
Carrington Labs, a fintech infrastructure provider focused on credit risk analytics and cash flow underwriting, launched an MCP server:
Carrington Labs’ MCP server delivers both speed and confidence to lenders investing in automation through two distinct capabilities. First, AI agents can access Carrington Labs' compliant model outputs in real time to receive deterministic credit risk assessments, unlike the probabilistic responses generative AI alone would offer, allowing lenders to build end-to-end automated workflows without reverting to manual processes. Second, credit officers can use large language models connected to the MCP server to ask questions in natural language and receive clear explanations of complex risk factors.
So what?Copy anchor linkCopied
What’s interesting to me about this news is the framing.
Lots of companies are working on figuring out how to incorporate large language models (LLMs) into their lending workflows. The value of refactoring those workflows to be more agentic is both obvious and highly compelling (I will have a deep dive essay on this subject coming soon 👀).
However, the approach most companies seem to be taking right now is to figure out how to infuse some LLM-powered capabilities (customer communication, parsing unstructured data, etc.) into lenders’ existing, non-LLM-powered workflows. This makes sense because, today, those workflows are almost all exclusively built on top of deterministic, rule-driven systems.
But that’s not how Carrington is framing its MCP server:
AI agents can access Carrington Labs' compliant model outputs in real time to receive deterministic credit risk assessments, unlike the probabilistic responses generative AI alone would offer, allowing lenders to build end-to-end automated workflows without reverting to manual processes.
Carrington clearly sees a near-future in which lenders’ workflows are primarily probabilistic, powered by AI agents. And in those workflows, the one piece that lenders (and regulators) will still want to function deterministically is the credit risk models used to assess the applicants’ eligibility. It’s that piece that Carrington wants to provide, via an MCP server integration.
I’m not sure we will ever get to that point (there are other parts of the loan origination and credit risk decisioning process that benefit from the structure and predictability of deterministic systems), and if we do, it won’t be anytime soon.
But credit to Carrington for thinking different.
#2: Zest AICopy anchor linkCopied
What happened?Copy anchor linkCopied
Zest AI raised some money:
Zest AI … today announced the successful completion of an oversubscribed, customer-led financing round. The investment was led by five of the company's key customers – SchoolsFirst Federal Credit Union, Members 1st Federal Credit Union, ORNL Federal Credit Union, Truliant Federal Credit Union, and Citi, through its investing group Citi Ventures.
The financing represents a significant valuation increase from Zest AI's previous growth round, reflecting the company's strong momentum as more financial institutions evolve legacy underwriting systems with AI-driven solutions. The new capital will power the expansion of Zest AI’s automation initiatives across the full borrower journey and drive wider implementation of its Generative AI-powered lending intelligence platform, LuLu.
So what?Copy anchor linkCopied
I’ll admit to being a bit confused by Zest.
The company was founded in 2009, which is ancient in fintech terms. It started as a direct-to-consumer subprime installment lender using machine learning to price credit for underbanked borrowers. It began exiting that business in the middle of the 2010s, after receiving regulatory scrutiny, and fully pivoted into a B2B AI underwriting platform for banks and credit unions in 2020.
Across those two eras, the company has raised a staggering amount of capital. A Series A ($19M in 2011), Series B ($23M in 2012 + a later $5M extension), Series C ($20M in 2013), and an undisclosed strategic investment from Baidu in 2016. Then, after fully pivoting to infrastructure, the company raised $15M led by Insight Partners in 2020, followed by a customer-led round totalling $18M in 2021, $50M from CMFG Ventures and Insight Partners in 2022, $200M led by Insight Partners in 2024, and now this undisclosed-but-oversubscribed round from customers.
Factoring in the undisclosed rounds, that’s likely more than $350 million raised over 14 years.
And what has that money bought?
A lot, according to the company:
With more than 50 issued and pending patents and over 650 proprietary credit models, Zest AI’s technology is used by nearly 300 lenders – from credit unions and community banks to large enterprise financial institutions – with solutions spanning automated underwriting, LuLu lending intelligence platform, and Zest Protect for fraud detection.
Of course, patents don’t generate revenue (unless you aggressively use them as a sword), and we don’t know how many credit decisions Zest is powering for those nearly 300 lenders (and how those decisions are monetized by Zest).
I will note that I am generally suspicious of late-stage customer-led funding rounds, especially when the terms aren’t disclosed.
Why does Zest need money, given that it raised $200 million from Insight Partners last year? And why wasn’t Insight willing to lead this round, when it had already doubled and tripled down? Did Zest incentivize customers to invest in this round (and the one in 2021) through the use of warrants or other mechanisms to align incentives?
I have no idea! I suppose we will find out when Zest eventually goes public or is acquired.
#3: CoinbaseCopy anchor linkCopied
What happened?Copy anchor linkCopied
Coinbase weighed in on the Treasury Department’s GENIUS Act ANPR:
Crypto exchange Coinbase Global has urged the U.S. Treasury Department to keep its forthcoming rules for the GENIUS Act tightly aligned with congressional intent.
In a detailed response to the Treasury, Coinbase noted that Treasury should avoid imposing requirements beyond what the statute explicitly commands, warning that overreach could stifle innovation and undermine the law's goal of making the U.S. the "crypto capital of the world."
So what?Copy anchor linkCopied
I wrote about the industry responses to Treasury’s ANPR in Friday’s newsletter, but Coinbase’s comment letter wasn’t available when I was doing my research.
As expected, the letter covers a wide range of topics, including the tax treatment of stablecoins, the scope of the GENIUS Act, and the definition of payment stablecoins.
And, of course, the question of yield-bearing stablecoins.
Unsurprisingly, Coinbase believes that the prohibition on offering yield or rewards should only apply to stablecoin issuers:
The statutory text is unambiguous: the prohibition on interest or yield payments applies only to payment stablecoin issuers, and it may not be extended to other entities on an “indirect payment” theory … The interest-payment prohibition thus does not, for example, prevent a digital asset exchange or other intermediary from paying loyalty rewards to accountholders, a benefit common to many types of business relationships.
But the best part is Coinbase’s argument for why Congress chose not to extend the yield prohibition to affiliated third parties:
[Congress] declined to include non-issuer third parties within that prohibition because banning other types of payments on stablecoins across the board would have inhibited growth and innovation of the stablecoin market — contrary to the GENIUS Act’s core purposes. By leaving distributors free to encourage stablecoin adoption and loyalty in the same way that credit card providers and other payment networks do today, Congress imbued GENIUS with a pro-market, pro-innovation philosophy and sought to promote the adoption of stablecoins in the United States.
This is an absolutely ridiculous argument to make. If Congress’s overriding goal with GENIUS was to drive the growth and adoption of stablecoins within the U.S. (where stablecoins have very little utility relative to other forms of money), why would it restrict any entity in the stablecoin ecosystem (including and especially issuers) from offering yield?
I mean, just think about the analogy that Coinbase is using to justify its argument: credit cards.
In the credit card market, the issuers of the cards are the ones that pay out the rewards. Imagine if the CARD Act had banned issuers from offering rewards, but Congress had forgotten to explicitly include non-bank co-brand partners in its prohibition. Would it be reasonable for Costco to offer rewards on the Citi-issued Costco Anywhere Card, especially if, in this hypothetical, Costco owned a major equity stake in Citi? Would that be aligned with Congress’s intent?
No, it obviously wouldn’t.
The most reasonable reading of Congress’s intent regarding the yield prohibition in the GENIUS Act is that it was intended to prevent payment stablecoins (emphasis on the word payment) from competing with yield-bearing bank deposits. You can take issue with that policy choice (though, personally, I do not … credit creation is very important!), but that’s the choice that was made.
2 READING RECOMMENDATIONS
#1: Meta is earning a fortune on a deluge of fraudulent ads, documents show (Jeff Horwitz, Reuters) 📚Copy anchor linkCopied
Internal projections at Meta show that as much as 10% of its 2024 revenue came from ads for scams and banned goods. That is a shocking number, even if the general trend — online platforms facilitating scams for which they are not financially liable — isn’t new or all that surprising.
#2: Tech Billionaire Marc Andreessen Bet Big on Trump. It’s Paying Off for Silicon Valley. (by Jake Pearson, ProPublica) 📚Copy anchor linkCopied
This is a good overview of how the dismembering of the CFPB has benefited Marc Andreessen and a16z, which is a relationship that deserves scrutiny, given Andreessen’s public comments regarding the agency.
*Bonus: Unlocking Pay by Bank’s Potential (by me, with Trustly) 📚Copy anchor linkCopied
Incumbents don’t panic without reason. JPMC's moves to charge for data access says a lot about what’s coming next. Pay by bank isn’t hype; it’s a redesign of merchant economics. See why the shift matters.
* this rec is brought to you by one of our fantastic brand partners
1 QUESTION FROM THE FINTECH TAKES NETWORK
There are a TON of interesting questions being asked in the Fintech Takes Network. I’ll share one question, sourced from the Network, each week. However, if you’d like to join the conversation, please apply to join the Fintech Takes Network.
What are the best arguments for a 50-year conforming mortgage?
My default position on this idea is that it’s not a very good one and that we should just focus on building more houses, but I want to stay open-minded. Why is this proposal good?
Please ensure that any arguments you make in favor of the 50-year mortgage are accurate, mathematically.
If you have any thoughts on this question, reply to this email or DM me in the Fintech Takes Network!
Thanks for the read! Let me know what you thought by replying back to this email.
— Alex
Join 2,444 other finance and fintech leaders in the Fintech Takes Network
The conversation doesn't have to stop here
Keep learning and connecting in the Fintech Takes Network
EVENTS | FEED | LIBRARY | DIRECTORY
