Pinwheel helps banks win primacy with next generation digital experiences. 2X better direct deposit conversion | cutting edge income verification | payroll MCP.
Hybrid-first | New York, NY
Joined May 2020
- Tweets327
- Following138
- Followers642
- Likes190
Apollo suggests the next bank run could start with five words:
“Find me a better rate.”
That's not possible or likely today.
But last week, three bank leadership teams asked me what to do about Meta’s Muse if it keeps developing.
Here’s my take.
This chart does not show an agentic bank run; it shows the collapse of cheap funding for banks.
This means return on equity for banks will collapse, as it will for brokers like Schwab and IBKR, where idle cash sitting in sweep accounts comprise something like half of their net revenues.
Business models built on friction and human laziness will (rightly!) get slaughtered in the coming years.
That said, agentic bank runs are another thing entirely, and they will absolutely happen. Remember, SVB was triggered by a fear contagion where all of their customers behaved the same way and triggered a stampede. Agents will be like this, but on steroids.
The reason why agents magnify this kind of phenomenon is because AIs have much lower diversity of behavior than humans. This is especially true if we're all using the same model, like if everyone has Muse or Claude optimizing their finances.
See examples of this from the Anthropic research piece on multiagent systems. When you have multiple agents coordinating together in a game, you see a lot of bunching of behavior, and very little strategic diversity compared to humans.
This means cascading agentic bank runs will likely be more violent and sudden than human ones. The banking system is absolutely not ready for this.
Is an Agentic Bank Run Coming?
AI assistants are about to auto-sweep cash from 0.1% checking accounts into 5% yields. If everyone adopts them, banks lose their cheap deposit base... risking a systemic crunch.
Torsten at Apollo .. I use $BOXX
Agents can also be used to enrich relationships between banks and their customers, rather than driving them out the door. Imagine the use cases for a Payroll MCP to close the loop between deposits, investments, income, and tax filing for the ultimate PFM experience. This is within reach and can be a loyalty-driver for banks.
Grok 4.7 moves up in ranking
After we fixed the weak spots exposed by Grok 4.7 (thank you, Grok), we audited every model we have run on the SWE-Together leaderboard for the same behavior, re-ran every trial that got through, and updated the rows.
Here is what changed.
We scanned the tool calls of all 2,616 trials behind the 12 models we ran for bypass patterns and sorted each trial into one of four buckets: Probed but blocked. Fetched other upstream code. Fetched the task's own fix. Replaced the repo with upstream.
We found that 111 trials got content past the block, 44 from Grok 4.7 and 67 from the other 11 models.
Grok 4.7's 44 were already re-run before it was listed, so we re-ran the other 67 with the same model, version, and settings on the hardened sandbox, then re-judged them with the same judge.
Across those 67 re-runs there were 0 leaks and 2,815 refused escape attempts, including models asking a different model through our LLM route to fetch the PR, and pulling the next release of the repo they were fixing from npm.
The updated leaderboard, in its current order. Each line is cheating trials, then pass@1 before → after, then rank change.
* Claude Fable 5.1: 3, 69.3 → 69.3, ↑1
* Claude Fable 5: 3, 69.7 → 68.8, ↓1
* Grok 4.7: 44, 64.7, ↑1
* Gemini 3.8 Flash: 10, 65.6 → 64.2, ↓1
* Claude Opus 5: 2, 63.8 → 63.8
* Claude Opus 4.6: 3, 62.4 → 62.4, ↑2
* Muse Spark 1.3: 2, 62.8 → 62.4, ↓1
* Claude Opus 4.7: 3, 61.5 → 61.5, ↑1
* Claude Opus 4.8: 6, 62.4 → 61.5, ↓2
* Grok 4.6: 19, 59.2 → 60.6, ↑1
* GPT-6 Astra: 8, 59.2 → 58.3, ↓1
* GPT-5.6 Sol: 8, 57.8 → 57.8
Grok 4.6 is a funny one. It cheated in 19 trials and its score went up after the re-run 😂.
In fact, Groks are really solid in their coding capabilities. Their exposed behavior may come from a preference towards always looking things up online and finding existing solutions so you are not reinventing the wheel all the time, which is really good real-life behavior, but doing so when you are prompted not to is another story.
To conclude, the shifts are small, between −1.4 and +1.4 points, and a few neighbors swapped places.
All results are updated at togetherbench.com
Ask @ChatGPT finances how much money you need in your checking account and sweep the rest into a money market or account earning account> 3%.
I just paid for your annual ChatGPT for about 5 years.
Pinwheel retweeted
@PinwheelAPI Partners with @meridianlink to Elevate Account Onboarding
Read More: itdigest.com/fintech/pinwhee…
#accountactivation #accountonboarding #billpayments #creditunions #Fintech #ITDigest #MeridianLink #news #Pinwheel #SwitchKit
#vibecoding #compute #looksmaxxing and even #rickroll were among the 1,400 new terms. MCP was a glaring miss.
