Things to know 👀
Erebor is the national bank Palmer Luckey and Joe Lonsdale built to fill the hole SVB left, and it opened in February. The FT reports it added more than $3bn and 500+ customers in the past quarter, taking deposits past $7bn, and it is raising $1.5bn at an $8bn pre-money valuation. Deposits went $1.1bn in March, $4bn in the June call report, $7bn now. 500 new customers for $3bn is a $6m average account, so most of this money sits above the FDIC's $250k cover.
🧠 This is very clearly a bank attracting wholesale deposits. This was the goal: solve for corporates who risked being de-banked in defense or crypto. The challenge is that wholesale deposits tend to move quicker and attract less of a “liquidity coverage ratio” than retail deposits, which are considered stickier.
🧠 The deposits are likely forcing the raise. Erebor has to hold a 12% leverage ratio, so a $7bn balance sheet needs roughly $900m of capital and it had $597m at the end of June. Larger banks tend to have a lower imposed leverage ratio (although they hold around similar levels for safety). Erebor’s raise is helping it keep within that guardrail.
🧠 Can they keep raising forever as deposits flood in? Raising VC is expensive. At some point, the need to generate revenue beyond just holding deposits and moving value. Palmer Luckey is incredible at raising, but if you keep selling equity, the return on equity (ROE), starts looking pretty poor pretty quickly.
🧠 Erebor offered stablecoin redemption at par with no fees. At launch, Erebor converted USDC and USDT to dollars at $1 per token, for free, while Tether charges the greater of $1,000 or 0.1% to redeem directly. Solving these “mint/burn” fees is huge for anyone in the industry.
🧠 And then the market makers arbitraged them. Wintermute and Galaxy Digital bought USDT a fraction below $1 and cashed it at Erebor for the full dollar (per The Information). Erebor told them to stop, kept free conversion for big depositors, and put volume fees and limits on everyone else.
🧠 Will they ever start lending? Is that even in the plan? Or is there another fee-based path to revenue? I guess we’ll find out when the Q3 call report drops at the end of October. It will show how much of the $7bn is cash at the Fed and how much is lent out (the March report showed zero loans).
🧠 I can't think of a de novo that has grown like this. No doubt that’s a sign of huge corporate demand for a better wholesale banking partner with good tech. But also, anyone who’s been in the industry for a long time will tell you, deposits that arrive quickly can leave quickly too.
Rain, the Visa and Mastercard principal member behind roughly 75% of tracked stablecoin card volume (per Dune), applied to the OCC on 5 October to open Rain National Trust Bank. If approved, the bank would custody digital assets and USD as a fiduciary, manage reserves for permitted stablecoin issuers under the GENIUS Act, and issue and redeem its own USD stablecoin as issuer of record. Rain was valued at $1.95bn in January and grew its active card base 30x last year.
🧠 Rain is arguably the growth “fintech” story. It has both network memberships, most of the card volume, payouts in 50 currencies, two loyalty acquisitions, and now a filing to custody and issue. Its clients like Kast are able to quickly add new markets and generate higher interchange than traditional domestic-only, debit-based programs. That economics advantage could grow with a national trust charter too.
🧠 The issuer keeps the income on the reserves. Today that goes to Circle, Tether or whoever issued the coin on the card. Rain would keep it as issuer of record.
🧠 This allows them to do some interesting new co-brand products. Uptop already runs rewards for the Cavaliers, the Pistons, and LSU Athletics. Rain bought Uptop in November and Ansa in August, so loyalty, stored value, and the coin can sit on one stack.
🧠 Imagine a sports brand with its own stablecoin and stablecoin-linked cards. The card works anywhere cards are accepted, and the stadium POS applies the discount, even through Apple Pay. The brand keeps a share of the reserve yield and the interchange.
🧠 Rain hired a bank CFO to run the new bank. Brandon Soto was CFO of Square Financial Services, Block's Utah industrial bank, and then CFO of Coastal Financial, a major fintech sponsor bank. There's a real war for senior banking talent to run trust charters right now.
🧠 The stablecoin boom might not last forever. The ICBA sued the OCC on 2 October to stop crypto trust charters. Its complaint counts at least 21 approved or conditionally approved, 13 of them crypto. Rain filed three days later.
🧠 There’s also still a risk of wallet attacks and hacks in anything on crypto rails. For example, an attacker drained about $1.1m of card balances at Avici and Tria in late August through an outdated Rain contract on Solana. Rain's pitch for the bank is a custodian "that answers to a federal regulator." Expect the comment period to ask about August.
🧠 The OCC still has to approve it, with a public comment period, and the bank would be uninsured. Rain calls it a multi-year project. If it works, one company issues the card, holds the balance, and mints the coin. That is something only Stripe is currently positioned to do today.
The FT reported that Wise contacted around 4,000 users of its Wise Asset investment service to inform them that they had been provided with incorrect tax statements between 2021 and 2025. The error came from miscalculations made by a third party software provider causing users to input incorrect data during their tax filling. Wise will bulk settle with the HM Revenue tax authority and compensate customers.
🧠 Wise is having a string of bad news. In June, reporting broke that Wise was being investigated over its anti-money laundering conditions by Belgian authorities. Then money laundering concerns also led US regulators to reject Wise's application for a US banking licence.
🧠 This doesn’t appear to be impacting growth. Net revenue reached $714.0 million in Q1, up 25% year-over-year compared to $573.3 million in Q1 FY2, with customers up 21% YoY. The core engine is still their consumer side cross-border business.
🧠 Wise is a solid default multi-currency account for businesses, with cards. This year the company as a 38% increase in corporate card spending across North America, but this is a segment where Airwallex is becoming increasingly competitive.
🧠 My experience of the UX in Wise is still poor. There’s little frictions everywhere in things like managing accounts integrations, or managing team cards. The stuff they’ve stretched into still feels half baked, which is an opportunity for others.
🧠 But whenever a company gets a difficult phase like this. It can be the making of them. I remember writing a few years ago about Stripe’s difficult teenage phase, and now look at them.
4. Meta and Sierra launch Personal Agent Protocol, a standard for how AI agents sign in to businesses
Sierra and Meta announced Personal Agent Protocol (PAP) on 6 October with Genesys, Instinct, Rocket, Shopify, Stripe and Walmart. It gives a personal agent like Meta's Muse a standard way to sign in to a business as the customer. The customer chooses read-only or write access, the business chooses whether the agent comes through its website, its APIs or its own agent, and the v0.1 spec is due later this month.
🧠 Every payments company, bank and merchant should pay attention to this protocol. There's a "what do we do about agents" question out there, and this is the first attempt at an answer with wide-scale distribution baked in.
🧠 Today a personal agent uses your website the way a human does. It clicks through pages, fills in forms, and phones support when it gets stuck. The business often has no idea an agent is in the session.
🧠 Amazon's answer has been to keep agents out. It is actively blocking services like Perplexity and Muse, saying they break its terms of service. And while they can create some fraud risks, in Amazon's case I suspect it's more about their $68bn in advertising revenue. (See Amazon vs Muse)
🧠 PAP is the industry version of a rulebook with agent sign-in and distribution. Out of the gate, Muse and Instinct is a great starting point. For the two agents with the most adoption, a business can see which agent is acting for which customer and set rules for what the agent can do.
🧠 It starts on a website, where an agent learns what a company offers and whether it's PAP-compatible. Then it starts a session for the user, and when it needs access to a customer account, the agent can sign in using credentials the user gave it. The customer gets to decide if the agent has read-only or write access.
🧠 Wallets are the next part of this story. Stripe Link, Shop Pay, and PayPal are appearing in Muse and Instinct. Those wallets will hold the credentials and be increasingly critical to how users manage permissions and payments.
🧠 Meta's Muse and Instinct are the only personal agents on the list currently. But Sierra CEO Bret Taylor says he expects OpenAI and Anthropic to join. I just wonder if he can convince the holdouts at Amazon.
5. Exclusive: Walmart, Ribbit-backed OnePay launches a ChatGPT Plugin
Walmart’s fintech app OnePay has enabled its 7.5 million monthly active members' data to be accessed through an official app inside ChatGPT across all paid and free tiers. The data is also available now inside the ChatGPT Dots, OpenAI's new always-on agentic assistant. The compay shared in an email to me: “Connecting to an MCP server manually is a technical task, and we built it for early adopters. Today we're closing that gap by launching an official OnePay plugin in ChatGPT and Dots.”
🧠 Why’s this news? It’s just a plugin. There’s a bigger picture here. Generally, I tend to agree with you. A company doing an AI thing isn’t news. But I think the timing of this is interesting. Companies are seeing success with ChatGPT plugins, and it’s a way to engage beyond running an MCP server.
🧠 The personal agent revolution is here. Muse and Instict are live. They’re early but scaling, and expect responses from other big tech companies. There are two possible reactions you can have to that. 1) Try to block it or 2) Lean in and try to figure out how it's useful for customers. OnePay has chosen the second path.
🧠 OnePay is an interesting animal. It often gets overlooked, but with 7.5 million members, it's becoming a meaningful player in the ecosystem and is incredibly ambitious. Where mainstream Neobanks have shied away from doing much with personal agents, I think we needed someone to be more experimental.
🧠 This is a pattern for other banks to follow. If you want to understand how to play in the personal agent ecosystem, job number one is to show up. Start with read-only and build out from there as you build confidence.
🧠 Hasn’t Plaid already integrated with ChatGPT? Yes. And on the surface, the two capabilities are the same today. But, if and when OnePay adds more data-write transactions- or starts giving accounts to agents, it could have a broader array of features than a single Plaid integration gives you.
🧠 In time, I think you want both. Your agent should be able to see all of your accounts, but have scoped, deeper access to a daily driver account, and the ability to sweep funds across them. The question is, will you be that daily driver?
4 Companies 💸
1. Fin.com Local payouts for global payment companies on stable rails
Fin lets payroll providers, money transfer apps and crypto exchanges pay people into local bank accounts across more than 30 countries. Customers can send stablecoins and have Fin convert them into the currency the recipient needs. The company reports billions in payment volume and says it is already profitable.
🧠 For a payroll platform, the payment isn’t finished until the worker can spend the money. A lot of stablecoin solutions deliver a stablecoin and expect the end user to keep it or figure out how to off-ramp it. Fin follows Latitude, which also recently raised in connecting local payment methods to stablecoins in one place. I just wonder whether these companies have capped upside. Are they better suited as an acquisition for one of the bigger stablecoin alrounders now, like Mastercard (BVNK), Bridge, etc. Fin has been acquisitive too, so maybe its them.
2. Pave Finance Personalized portfolios for wealth advisors
Pave helps wealth advisors build and manage investment portfolios around each client’s circumstances. Advisors can account for shares a client already owns, their tax situation and preferences such as excluding particular industries, while automating the work of keeping portfolios aligned with those requirements.
🧠 Reducing the intake and servicing cost could help advisors take on younger clients years before they become wealthy. I wonder whether the result is lower minimums and more people getting advice, or just better margins for the existing firms. There are also a lot of companies playing in this space right now; it's hard to tell which is getting any real traction.
3. Integral Accounting, payroll and taxes handled for small businesses
Integral helps small businesses get their bookkeeping, payroll, and tax filings done through its affiliated accounting firm, Integral Tax. AI prepares the work, while licensed professionals review it, handle exceptions, and take responsibility for the filings.
🧠 Tax prep and submission is about to be majorly disrupted. Integral says its professionals can now serve twice as many clients each, which is a much more useful measure of AI adoption than how many employees use a chatbot. Nobody likes tax prep, especially in the US. Using AI helps, but you want to know it's right. Putting those two things together is smart. I’d pay for this if they worked in the UK.
4. Qupital Working capital for e-commerce sellers
Qupital helps e-commerce merchants finance stock and other business expenses using sales data from marketplaces including Amazon and TikTok Shop. Sellers connect their stores so Qupital can assess the business and offer financing. The company reports more than $9.5bn in cumulative loans processed and profitability over the past two years.
🧠 All lending is being rewired with live sales data. Sales data gives a lender a way to understand that demand even when the seller’s bank balance looks unimpressive. Again with these things though, why this, and not the 30 others?
Good Reads 📚
A record 82% of Brazilian households are in debt, and a third are behind on payments. The IMF's July review says credit cards and unsecured personal loans drove much of the build-up, with fintechs and digital banks leading the lending. Reuters opens with a public servant earning about minimum wage. Nubank raised her card limit twentyfold over three years. She now owes five times her monthly pay and has moved in with her mother.
🧠 Brazil has a debt problem: Revolving card rates are about 440% in Brazil vs about 21% in the US. Neobank card delinquency rose from 7.71% in 2021 to 20.31% in 2025. At traditional banks, it fell from 14.57% to 13.6%. More than half of Brazil's 96 million cardholders carry debt, and payments take 54% of their income on average
🧠 Nubank’s numbers still look good even in that climate. Well below the Neobanks or the large banks. Their late-stage delinquency rate (loans more than 90 days past due) stands at 6.9%, while early-stage delinquencies (15 to 90 days past due) are at 4.8% in Q2.
🧠 This is becoming a political hot potato: Lula and Flavio Bolsonaro have both proposed tighter rules on credit offers ahead of October's election. Mercado Crédito and PicPay deny adding to over-indebtedness and say they raise limits based on payment behavior. Nubank says its Q2 results show disciplined lending.
🧠 Brazil has been a perfect storm: Pix brought a massive population into banking, in a market with a high base rate for lending (14%). First-time borrowers got high limits with few affordability checks, and the delinquency data shows the result.
🧠 I still think financial inclusion is the best thing fintech has done, and it can be done right. Brazil has the Open Finance data to underwrite on real cash flow, and the lenders who use it will be in better shape when the new rules arrive. Lending is easy; getting paid back profitably over the credit cycle is another thing entirely.
Tweets of the week 🕊
That's all, folks. 👋
Remember, if you're enjoying this content, please do tell all your fintech friends to check it out and hit the subscribe button :)
Want more? I also run the Tokenized podcast and newsletter.
(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
