How I used ChatGPT Dots to build a public economic data portal in about 90 minutes—and why experience still matters.
On September 29, OpenAI introduced Dots, always-on AI agents inside ChatGPT. The next morning, I used one to launch fred.davidabernathy.com, a public website that makes housing, financial and economic data easier to explore.
About 90 minutes of build time. Less than 24 hours from announcement to a live product. Then I headed down to the LuxuryRealEstate.com Luxury Conference.
For me, that short distance between an idea and something people can use is the most consequential part of this release. Professionals who already work with AI can put a new capability to work almost immediately.
What makes Dots different
OpenAI describes Dots as agents that can keep working toward a goal without needing instructions at every step. Powered by GPT-6 Astra, each has its own cloud computer and can connect to thousands of apps through plugins. Users can follow the work, give feedback and reach their Dot through ChatGPT, Slack or Teams. OpenAI’s announcement
That changes how you approach a project. You can explain the outcome you want, establish standards and let the agent carry the work forward.
I had already spent time with Meta’s Muse. Even with that experience, I was surprised by how capable my Dot felt during this build. It kept moving toward the goal and turned the brief into a working site.
A better window into the forces behind real estate
The FRED Data Portal is built for buyers, sellers, investors and agents who want to understand the broader forces affecting the market.
Its underlying resource is FRED, the economic database maintained by the Federal Reserve Bank of St. Louis. FRED brings together data from many sources, giving users access to indicators covering housing, interest rates, inflation, employment and economic growth.
I organized the portal around three areas:
- Real Estate: Home values, housing supply and construction activity, with national, state and county views where available.
- Finance: Mortgage rates, Treasury yields and Federal Reserve policy indicators.
- Economic Data: Growth, inflation, employment and market indicators.
The goal was to make useful information easier to find and understand.
I also wanted the presentation to respect the data. Reporting periods should be clear. Revisions should be visible. Missing observations should remain missing rather than appear as zeros.
Those details matter when a reader uses a chart to inform a decision involving a home, an investment or a business.
Ninety minutes of building, years of preparation
At about 7:00 a.m. Mountain Time on September 30, I gave my Dot the goal: build a public site that makes FRED data easy to explore across real estate, finance and economics.
Before 8:30 a.m., the site was live.
The Dot did most of the build. I helped establish the FRED API connection and entered the API key myself during the process.
My most valuable contribution was the brief: which information matters to a luxury buyer in Palm Beach County, how to organize it, and what a reader needs to understand before drawing a conclusion.
Eighteen years in real estate and a career in wealth management shaped those decisions. AI accelerated the execution. Experience gave the project direction.
OpenAI also makes clear that Dots can make mistakes and that consequential work needs review. That remains part of the job, even when the building happens quickly. OpenAI’s guidance
The advantage comes from practice
A new AI release gives people another tool. Daily practice helps them recognize what to do with it.
In my experience, moving quickly depends on three things:
A clear problem. I already wanted a simpler way to put useful economic data in front of clients and colleagues.
Domain judgment. You need to know which questions matter, which information belongs on the page and whether the result makes sense.
Practice with AI. Regular use teaches you how to explain a goal, spot a weak result and give feedback that improves the work.
Access matters, and new products roll out unevenly. But once a tool is available, preparation can make the difference between experimenting with it and producing something useful.
I have seen this pattern before
On October 6, 2025, I attended OpenAI’s DevDay with Cory French, whom I work with. As far as I knew, we were among very few real estate professionals there, alongside the Zillow team.
That day, Zillow introduced its app inside ChatGPT, allowing people to explore homes and rentals through conversation. Zillow described it as the only real estate app available in ChatGPT at launch. Zillow’s announcement
It reminded me of an earlier shift in my own business.
About 18 years ago, I went deep on search engine optimization. Our listings and community pages at Waterfront Properties earned strong visibility on Google. Sellers who owned businesses saw those results during listing presentations and asked how they could achieve something similar for their companies.
Learning the tools early created an advantage that lasted.
AI is compressing that learning-and-building cycle. A project that once needed a longer development process can now take shape in a morning. Knowing what to build still takes thought.
Start with a question people already ask
You do not need a grand AI strategy to begin. Start with one recurring question from your clients or customers.
What information would help them? What would make the answer clearer? Could you turn it into a resource they can return to?
Give an agent that goal. Review the result. Make it useful enough to share, then improve it as you learn.
That is what I did with the FRED Data Portal.
One morning produced a new resource. The experience behind it took years to develop.
As AI makes execution faster, that experience becomes more valuable: it helps you decide where to aim.