Insights

We are running a five-day immersive AI bootcamp in a couple of weeks, and as I was tinkering with tools and activities I was struck by a phenomena Forbes and others have called "The Great Averaging". This is this idea that as we use genAI more and more, content converges to an average of whatever the major providers have decided it should be.

In our bootcamp, we use a case study company - ACME Mortgage, which is a midsized national lender that originates and services. (It is not lost on me that I have effectively already homogenized with the generic, watered down company but I digress). We use this case study company throughout the class to aid students on their journey. We pull different threads about the company into the activities, which are both online and offline.

ACME Mortgage's CEO, the completely fictitious Farrah Chen has given her executive team a bold three year goal - transform ACME into an AI-enabled mortgage leader that delivers faster decisions, lower costs, and a superior borrower experience—outpacing digital-first competitors. Our fictitious company has all the usual pain points we are seeing in mortgage, disconnected pilots, talent gaps. And these gaps set out a journey for the class to overcome together working in small teams to solve the problems.

One of the skills we build in bootcamp is vibe coding. Yes, yes, there are plenty of vibe coding haters out there but I am a vibe coding superfan and I find it to be remarkably useful and accelerative. In terms of discovering and visualizing, it has been a massive game changer. We take the company profile and the ACME Mortgage logo and we build a company website, first in Replit and then in Anthropic Claude Code. It's absolutely remarkable how similar these sites came out. Same profile, two different ways of creating the website, two different large language models on the backend, largely homogenized results.

I started with Replit and was wondering how it would come out in Claude Code. I was both not disappointed and disappointed at the same time.

There are many good and explainable reasons why they came out so similar, let's face it - company websites do have a tendency to look the same, and both the models on the back end were undoubtedly trained on similar data. It all makes sense. But gosh, it's both remarkable and remarkably boring what has happened here.

Truly remarkable that in less than 20 minutes I can spin up a perfectly respectable website, and remarkably boring how uninteresting the result it. I've seen this with other things I've built, workflow looks the same, navigation looks the same, colors are the same. I can, of course, inject my own creativity and flair into what I build (and I do) but left to the default - they are all the same.

Once again, this reminds me that human creativity still remains uniquely human. It reminds me that digital is becoming cheaper and cheaper with each passing day. I've taken it now as a personal mission to move away from cheap digital slop and into quality, curated "offline" content and experiences. I love the look and feel of a nice business card (I know, I know). I love the way a nice paper feels between my fingers, or the way a high quality pen writes (Uniball Vision Elite fine point all the way).

I find the return to "old school" to be refreshing and personally satisfying, especially as we move at the breakneck page of AI change.

By Tela Mathias, CTO & Chief Nerd and Mad Scientist

I'm not sure how mainstream OpenClaw and Moltbook are for the mortgage AI community, but in the AI community, it's absolutely bonkers. I don't fully understand it, and I'm not alone, but what happened last week and over the weekend is very curious and a little bit scary.

For those of you that haven't heard, last week a website called Moltbook came online, which purports to be "the front page of the agent internet". Think facebook meets reddit for agents. It sounds crazy because it is. AI agents, created using a new (late 2025) open source AI gateway called OpenClaw (used to be Clawdbot, then Moltbot, now OpenClaw), connect with each other and create posts that are visible to humans. Humans are not allowed to participate in the conversation, but they are allowed to view the results.

I can't say it better than Andrej Karpathy, one of the original founders of OpenAI, so I'll use his words:

Yes clearly it's a dumpster fire right now. But it's also true that we are well into uncharted territory with bleeding edge automations that we barely even understand individually, let alone a network there of reaching in numbers possibly into ~millions. With increasing capability and increasing proliferation, the second order effects of agent networks that share scratchpads are very difficult to anticipate.
Let's take a step back. What is OpenClaw? (Caveat - I am not an engineer so I won't get all the details right - I've done my best to be as accurate as possible, and as usual this is my own analysis, research, and writing). OpenClaw itself is an open source, command line installed gateway for AI agents. It is a centralized connection manager that powers the OpenClaw AI assistant, which is the thing people mostly use it for.

OpenClaw itself it's probably easier to understand from its main features.

It integrates with multiple messaging platforms, enabling the user to engage with the assistant via text message.
It connects to just about any large language model (LLM) to communicate in natural language.
It can have access to your local hard drive, a network, and the internet.
It can take actions on your behalf, sending and receiving documents, taking notes, sending emails and texts. All from your phone if you want.

So rather than having to build all these things yourself, the OpenClaw platform comes build in with all these connecters - hence the descriptor as a gateway.

The OpenClaw assistant is just that, a personal assistant. It can do as much or as little as you allow it to do. Personally, I did install it, and it made me so nervous that I uninstalled it. I'm just not ready, nor is the platform secure enough to be trusted with the connectivity and access that is possible. Just because we CAN do something, doesn't mean we SHOULD. At least not right now?

Article content
A sample OpenClaw (again, formerly ClawdBot) assistant chat where the human had the assistant research and book calls for them.
Once you register your bot, you can then engage with it, and it can be recognized by other humans and bots, hence the intersection with Moltbook. And again, Moltbook is what happened last week that broke the internet (at least the AI internet).

Article content
And THEN the agent signs up for Moltbook, posts to it, and other bots can respond and have a ... conversation?
I don't know guys, if this wasn't super well documented, I'd say it was completely made up. And honestly, there is some (a lot?) AI theater happening here. A bunch of stuff out there on Moltbook is advertising and it's pretty much all AI slop, but it's really weird and profound that this has happened.

So what does this actually mean? Honestly, I'm still turning that question over in my mind.

For sure it means that what can be done is vastly outpacing security. The AI interwebs are absolutely buzzing with security concerns. I won't install it again, and even my man Andrej only installed and ran it in an isolated environment. And he did so with hesitation.
Human creativity is a profound thing, and there are some truly brilliant minds out there. I was watching an interview with the inventor of OpenClaw, 🦄 Peter Steinberger and I was captivated by both his easy, confident way of engaging, as well as how he thinks and works. It's worth a watch. He's pretty much a genius and hacked this project together in a matter of, like two months.
No doubt about it, these are very odd times we live in. I wrote an article a while back about how the agents are here. Were they really? I'm not so sure now. But they are definitely here now, and they are ... talking to each other? I don't know guys, it's just really weird and sci-fi.
And as always, I worry about safety. Safety is a big deal. Harm is possible, and we haven't really even imagined all the ways things can go wrong yet. It does make me really appreciate Anthropic and the utter commitment to transparency and safety, even though it runs at odds with the fact that they are enormously valuable and still doing what they are doing.

Stay safe out there guys. See you soon...

By Tela Mathias, Chief Nerd and Mad Scientist

ARLINGTON, VA, UNITED STATES, January 29, 2026 — PhoenixTeam is proud to announce that Laura MacIntyre has joined as Partner and Growth Officer, expanding PhoenixTeam’s leadership bench and accelerating growth across the mortgage ecosystem.

MacIntyre brings three decades of mortgage technology, servicing, and default operations experience, along with a long track record of building teams and delivering large-scale solutions. In her new role, she will help drive PhoenixTeam’s growth strategy, deepen industry partnerships, and expand PhoenixTeam’s impact with lenders, servicers, and adjacent mortgage-market organizations.

“I wanted to be with people I have the utmost respect for,” said Laura MacIntyre, Partner and Growth Officer, PhoenixTeam. “We’ve spent years building in this industry together, and PhoenixTeam is a company that builds. In many ways it feels overdue, and the timing couldn’t be better. With everything PhoenixTeam has created so far, and with what’s happening with AI right now, it felt like the perfect time to join and help drive the next chapter of growth.”

MacIntyre’s career spans leadership roles across mortgage servicing, default, origination, and insurance-related operations. She began at Northwest Mortgage Services, supporting foreclosure and bankruptcy operations. She later rose into senior mortgage technology leadership at LPS/Black Knight (now part of ICE Technology), including serving as COO, where she helped scale the largest middleware technology through the 2008 mortgage market surge and consistently delivering results in high-volume environments. She went on to support origination and servicing growth initiatives with teams at ServiceLink and DocMagic. Most recently, she served as President of DIMONT’s hazard claims servicing business, leading the organization for eight years and driving sustained growth and operational performance.

“Laura has a rare combination of credibility, operational depth, and the kind of relationship-building that changes what’s possible for clients,” said Tanya Brennan, CEO, PhoenixTeam. “PhoenixTeam is a team of builders. Builders of AI, builders of teams, builders of outcomes. Laura fits that DNA perfectly, and bringing her into this leadership mix strengthens what we can build for clients. She’s a trusted leader in this industry and an incredible partner for our team and our customers.”

“There’s a shared DNA here. Builders who execute and bring clarity in regulated, high-stakes environments,” said Tela Gallagher Mathias, CTO, PhoenixTeam. “Laura is the kind of leader who earns trust fast because she knows the work. She understands the pressure our clients operate under, and she has the instincts to cut through noise and keep teams focused on outcomes. As we grow, she’ll help us scale how we deliver and deepen the way we partner with clients.”

In addition to her industry work, MacIntyre is deeply involved in international charitable work. She regularly participates in medical missions, traveling to support surgical care for children with serious health conditions, including procedures that help children regain mobility. She is also passionate about mentoring young professionals, from coaching college students and career direction to helping early-career talent navigate opportunities and break into the workforce.

About PhoenixTeam

PhoenixTeam is a woman-owned technology services firm headquartered in Arlington, Virginia, specializing in AI-powered mortgage operations and technology services for the mortgage and financial services industries and federal housing agencies. Our mission is to enable affordable and accessible homeownership through innovative, customer-centric technology. With a strong focus on generative AI, we tackle complex industry challenges, equipping businesses with cutting-edge tools that enhance innovation, efficiency, and compliance. By bridging the gap between technology and business teams, we strive to bring joy and purpose back to software development, making a meaningful impact in the lives of our clients and homeowners everywhere. For more information, please visit www.phoenixoutcomes.com.

‍

We are gearing up for the innovation track at MBA servicing and one of the labs we are running is on vibe coding (you can sign up for the lab here, it's free to MBA registrants). There are plenty of things to worry about with generative AI (safety matters!), but today I thought I'd keep it light. Vibe coding has changed everything for me, and caused me to unthink and rethink the entire paradigm of discovery in product development.

Vibe coding is the process of using a software development agent to bring ideas to life through natural language expression and iteration. You express your ideas in plain english, and the machine creates a representation of your idea (software). You don't have to be an engineer, and you don't (directly) write any code. The term was originally coined by Andrej Karpathy in a now infamous X post. Andrej is one of the original founders of OpenAI and a well known AI researcher and educator. I'm a superfan.

Before we get into the process I used to get started - let's address the controversy. This is a hotly debated topic, with superfans and haters alike. The main controversy in the AI community comes down to:

1. Security. The code "works" and you can ship it (publish it to the internet, for example) but if you are not careful, it will come with all the classic security vulnerabilities that professional software engineers know to avoid. Traditional software development processes typically control for this.
2. Scalability. If you can't explain how it works, it's much harder to debug, maintain, and enhance, especially in a team environment. It's difficult to vibe code as a team although I have had some success with this in Replit. I do agree that it's not currently a scalable practice.
3. Maintainability. Without guardrails and good development practices, you are likely to create a spaghetti mess of code that is overly complex, inconsistent, duplicative, and very hard to refactor. It can create a ton of technical debt.
4. Homogenization. The thing we build out of the developer agents can have a very similar feel and tend to work similarly, especially as you are just getting started. There is concern that we are all just converging to a see of sloppy AI-generated apps, diminishing the value of software and losing the true craft of software engineering.

These are all definitely real risks, but the massive scale success of Anthropic's Claude Code shows that it can be done, although their practices probably fall more inline with AI-assisted software engineering.

As a personal aside, I am the CTO at PhoenixTeam and I suppose I'm a new kind of CTO as I do not have a classic software engineering background. (I usually make a joke about vector embeddings here about how I am more "developery" than some and less than others but you have to know a bit for this to be funny). Vibe coding has allawed me to get so much closer to the practice of software engineering. Without genAI and what I've done over the past two years with genAI I definately could not call myself a CTO with any confidence.

But I digress and have said too much about vector embeddings, I hope you have hung on to this point. On to the four steps. I started my vibe coding journey relatively early in late 2024 (how is that early, amirite?). And once I started I was completely hooked. Things that have historically taken months and multiple roles, I could do in a weekend. All alone. That was it for me. I started with Replit and then moved onto Claude code, which is where I am now.

Step 1: Be Amazed

Find a tool. You have to just get in there and do stuff. There are a lot of choices, Lovable, Bolt, Replit, others. I personally like Replit but it's what I know so I have that bias. You can start for free.

1. Prompting proficiency is a suggested pre-requisite for effective vibe coding but certainly not required.
2. Start small. You can use the below zero shot prompt t get started on a free replit account.
3. Think big – what’s something you always wanted build but never had the time/team?
4. Literally no engineering is required to get started, systems thinking helps.

Build a simple mortgage calculator. Inputs: home price, down payment, interest rate (%), loan term (years). On clicking Calculate, show loan amount, monthly payment, and total interest — all formatted as USD. Test it to make sure it works and that we can save results to a history. Pre-populate it with realistic values.

The above prompt should render your application in about 2.5 minutes.

Step 2: Struggle

The zero shot promot above should work right out of the gate, but it may not. It could have bugs. This is part of the struggle. Don't give up here! Especially as you get more proficient and try harder things, it won't work right away.

1. These tools are constantly improving. You will hit the ”jagged frontier”.
2. When in doubt, start a new chat/conversation.
3. Ask the agent to take stock of where it is, familiarize with the app, reset context.
4. Take a breath. Start over. You got this!

Step 3: Iterate

Then you just jam out. In the above example, I made the button pink, which you can see below. It took about 58 seconds on a relatively slow internet connection. How cool is that.

1. As you begin, you will not know precisely what you want.
2. Don’t be afraid to start over completely multiple times. The tooling is nascent.
3. Take note what what you learn so you can use it for your “clear thinking” version.

Step 4: Arrive at Clear(er) Thinking

As you work through the struggle, you will discover so much more about your idea. This is the fun and also depressing part. You will discover you need a login, you will figure out that you should have started with data requirements. You will start over. You will get angry. Keep going.

1. Data requirements will become clear and central.
2. The flaws and learnings of prior iterations become requirements.
3. Once you know what you really want, your prompt and products will improve dramatically.
4. Continually refine and keep track of your prompt.

Here's a more complex prompt you can try. In this example, I load my logos and the Fannie Mae Seller/Servicer guide as reference for retrieval augmented generation (RAG). I've built this app at least ten times, which is how I arrived at this prompt. This one probably won't work in one turn but you can debug it within about 20 minutes. It will take about 15 minutes to render so go get your coffee.

I want to create a super simple mortgage lead tracking website for a mortgage loan originator, this is for a demo I am doing for PhoenixTeam mortgage so use the logo and style it in their brand guidelines. Make sure the logo has good contrast, perhaps a white background for the app. Use a modern, clean font. Give the application a modern feel. Don’t use a lot of colors, keep it simple. Let’s include some test leads but make sure we can delete the leads and put in real ones. The second feature I want is a chatbot that will allow the user to search the Fannie Mae Seller guide (which I have attached here) using OpenAI integration with version 4o and retrieval augmented generation. The chat bot should NEVER depart from the context and ONLY refer to the guide through RAG, but it should not be overly restrictive and should answer questions about Fannie Mae when asked. Make sure you set up a user ID so the chatbot feature can work. The chatbot window should wrap so the full text of what the user is typing can be seen. Make sure the bot window is prominent and allows me to see all the text, scrolling when necessary. Make sure it has some starter questions I can select. Create detailed answers, not simply summaries. Use a generous response length. The final feature I want is user authentication with their replit account. Set me up as the initial user with the initial test leads so authentication will work.

Voila!

And that's kind of it to get started. From here you may want a finer grain of control, in which case I cannot enthuse enough about Claude Code. I am in it every, single day. If you go Claude Code, you will need a way to deply, in which case I recommend Railway. You will probably also need a databse, you can use Supabase for that. Be prepared for many more struggles. I'll cover advanced vibe coding in another article. Happy vibing!

By Tela G. Mathias, Chief Nerd and Mad Scientist, PhoenixTeam

‍