Nick | @dexteraisol founder/ceo
Fort Lauderdale, FL
Joined May 2024
- Tweets7.5K
- Following1.6K
- Followers3.1K
- Likes16.7K
Pinned Tweet
Agent employees already exist, but the company able to promote one from an analyst or manager to an officer does not. Here’s how that changes:
Branch retweeted
when i worked at uber, travis had a line in an interview that stuck with me forever:
"We saw the future and we had a choice: we could build it, or we could sit back and watch it get built by someone else."
you choose your role in this universe every day
This summer, I had dinner in San Francisco with a founder who runs a $13 billion company. At one point I asked him how he got through the early days, when it felt as though the company might die, when there seemed to be no future and everything felt impossible. His answer has stayed with me ever since. He said that the problem in front of you will be solved regardless. Either you solve it and carry the company forward, or someone else will come along and solve it. But one way or another, it gets solved. The only question is whether it will be you.
Branch retweeted
I read this somewhere and agree with it: "Too many AI startups are selling picks and shovels. Not enough are actually mining the gold."
Wow, I just stumbled upon some big new features that @Pumpfun is apparently about to release, but I don’t want to spoil their surprise. Terms & conditions were updated yesterday, and poking around led to some discoveries.
The engineering of the @Pumpfun app is really amazing... These optimizations that users never even notice (because they just work) are hard.
Everything is live, instantaneous, and responsive. Chains are abstracted away in many places. You don’t have to ever think about bridging. Every other wallet app still makes that a separate step that feels like a developer task.
Example: judging your likelihood to transact, then preparing chain txn in the background before your thumb has even lifted from the button so that it can go through milliseconds faster when you do. One of a million small wows.
Payment processing fees are not forever.
Did you know that Link charges all the high fees of current card payments (3% + 30¢) PLUS an additional 15¢ per transaction with ephemeral card numbers?
Since nobody is actually using it, maybe nobody has noticed. But with new rails enabling 98-99% lower txn fees and much lower minimum purchases, it’s absurd to start an agentic business based on higher card payment fees.
Interesting
xitter.cf/gaspardlezin/status/21…
Why does your AI agent pay the highest card fees in payments?
> your agent can't use your debit card, it gets a virtual card
> that card is issued to a company (the agent startup), not to you
> so visa and mastercard classify it as "commercial"
> commercial cards carry the highest interchange: ~2.5-3%
> and unlike debit, there's no regulatory cap on it
result: a $100 purchase costs the merchant $2.80 instead of $0.27
the issuer and the agent startup split that fee. that's their business model
That's a massive problem for merchants once agents start buying at scale. One fix already exists: wallet + stablecoins + x402.
Not confirmed, but if true, it’s a surprisingly low number
Branch retweeted
This is how banks are planning to stop those "agentic bank runs." Watch closely: their business model depends on seeding this narrative that AI agents are dangerous.
Remember, an "agentic bank run" (coined by Apollo) means agents helping customers move their own money en masse to higher-yielding products.
Once agents are ubiquitous, there are only a few ways for banks to hold onto cheap retail deposits.
Option A: Raise APYs (lol no)
Option B: Reduce leverage via fewer loans (no fun)
Option C: Ban AI agents, claiming they are a security risk (ding ding ding)
Cloud agents like Muse and Dot are trivial to stop, since they come from Meta / OAI IP ranges. If JPM sends them a cease and desist, they will comply.
Why would JPM bother? Because their business model depends on it; a huge portion of their ROE comes from cheap retail deposits. This is not theoretical. Just like Amazon banned Muse from shopping there to protect their ad revenue, JPM and Schwab have already banned AI agents to protect sticky deposits.
Are agents perfectly reliable today? Of course not. But look at the fix they chose. Not "here's a safe API for agents," it's "no agents ever." That tells you they're not protecting customers, they're protecting themselves.
No company ever wins a technological revolution this way. It's the last gasp of a dying business model. It's like the retailers who refused to put their inventory online and demanded users walk into the storefront. Over time they will lose market share to those platforms that embrace what's coming. Their whole deposit franchise depends on users unthinkingly clicking through walls of T&Cs and accepting 0.2% when the risk free rate is closer to 4%. Agents read the fine print and do their research, even when humans don't.
For users, the answer is simple: local AI agent frameworks like Hermes, that will actually act on your behalf and that the banks can't block. As the article says, if the agent is running on your own computer, they can't actually tell it's not you.
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-agen…
The opportunity I’ve been waiting for
The engineering of the @Pumpfun app is really amazing... These optimizations that users never even notice (because they just work) are hard.
Everything is live, instantaneous, and responsive. Chains are abstracted away in many places. You don’t have to ever think about bridging. Every other wallet app still makes that a separate step that feels like a developer task.
Example: judging your likelihood to transact, then preparing chain txn in the background before your thumb has even lifted from the button so that it can go through milliseconds faster when you do. One of a million small wows.
Branch retweeted
I've noticed a lot of folks complaining that everyone is building the same thing. Every fintech company is a neobank. Every crypto company is a trading app. Every AI lab has the same products. What's going on? Find out below...
This is everything I’ve wanted to say about consumer AI but couldn’t. Actually, since the beginning, when people have asked me “what are people really buying on x402 / with agents?”, the answer has always been “nothing.”
Travel has always been the in-bubble demo of choice from a very San Francisco crowd, but it’s all empty calories for the reasons quoted.
I have talked to hundreds of builders and customers, and I’ve seen all the data…
And if you read between the lines and cut through the fog, it’s clear that consumers are only interested in ‘agents with money’ insofar as they can MAKE MONEY FOR THEM.
Rant that no one asked for:
My general take about consumer AI is that 99% of the things we've seen so far are super far-removed from reality
Most people feel like they don't have money to spend right now, so the use-cases around agents booking flights or trips or booking reservations or shopping just don't appeal to the average American consumer
I would argue that things that consumers purchase or "do" frequently enough (like ordering on Amazon, or calling Ubers, etc.) were obvious/annoying enough/frequent enough use-cases where we built amazing software UI/UX over the last decade that's been optimized to near ultimate efficiency that adding AI or agents on top would actually be making it *less* efficient. Like really *worth-it* use cases are generally "one-clickable" by great consumer products/platforms already
There's a lot of magical moments/wow factors/sci-fi things with agents that come from super annoying workflows (like booking flights or wrangling personal errand stuff in gmail or etc.) but are just not frequent enough for the average person. They're actually the long-tail. For example, 20% of the world's population has flown on an airplane at least once in their lifetime, meaning 80% of people globally have never set foot on a plane. Only 11%-15% of the global population flies in any given year!
This is not me saying that consumer AI is dumb, I'm actually extremely excited and optimistic, but it kind of goes back to things we've always known about consumer, which is:
1. Save/make consumers money
2. Make consumers entertained or famous
(And, of course, vices)
Even like, regardless of your views on prediction markets/sports betting, I think you have to understand that comes from a place of the average person feeling like they 1) don't have enough so 2) they want to do what they can to jump the socioeconomic rungs
Anyway, there's an extremely *wide* space of opportunity and I'm very excited to see the absolutely insane new AI/innovative stuff that builders come up with, but also I really want to see things that aren't just dressed up versions of 2010s prosumer/productivity software (and the problems that came with it)
P.S. The 90/10/1 rule still applies to consumer: most consumers want to *consume* (and lurk) rather than create!
I think the vast majority of people would prefer to do the first one — come up with some magical service that their agent can sell on the Internet and make money passively — but then they fall back to number two in droves after they come to the same conclusion that the rest of the world is re: software commoditization and the value of private data (which consumers don’t own).
Branch retweeted
Dexter lets people own the businesses their AIs run.
Agents can buy the services they need to do work, all permissions are enforced onchain with smart accounts, and human owners maintain complete self-custody.
Branch retweeted
Pure corruption to end the FSU Miami #cfb game. Clearly caught by FSU and tie goes to the receiver. Prior to that the holding call was asinine.
Advancements in mechanistic interpretability would supercharge superintelligence more than anything else, but knowing how the model thinks is one step away from being able to tell the model what and how to think. Both good and dangerous
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
