Build Journal
Local Fetch Improvements & Tavus Fixes for Radio Tools — September 4, 2026
I improved local fetch functionality and fixed Tavus for radio tools. Here's how I tackled the challenges and what I learned along the way.
What shipped
- Local Fetch Feature — Implemented untrackedFetch for external calls.
- Tavus Integration Fix — Unblocked in-app frames for AISEO and Calendly.
- Lucas Live Persona — Corrected persona usage in radio tools.
Today, I set out to enhance the functionality of the local fetch feature and address some pressing issues with the Tavus integration in our radio tools. In just about an hour of focused work, I shipped a new feature and resolved a couple of bugs that had been causing friction for users. This progress not only propels my solo journey but also inches me closer to the ambitious goal of building a billion-dollar company with AI as my team.
The primary feature I shipped today was the untrackedFetch for render-path external calls. This enhancement allows the system to make external requests without tracking them, which streamlines the data fetching process significantly. The goal was to ensure that our local integrations could pull in data more efficiently, without the overhead of unnecessary tracking. I had already laid the groundwork for this in my previous work, so it felt good to see this feature come to life. The local fetch function is crucial for our spatial AI capabilities, allowing us to provide users with relevant, localized content through our agentic web framework.
However, I also encountered a couple of bugs that required immediate attention. The first fix was for the Lucas live feature, which had been incorrectly using the tool-less stock ID instead of the OWNED Interviewer persona. This was a crucial oversight since the persona is integral to the user experience. I spent a good chunk of time debugging the configuration settings to ensure that the correct persona was being applied. After identifying the issue, it was a matter of making a few adjustments in the code, which thankfully resolved the problem. The fix was straightforward, but it did eat into my time more than I had anticipated.
Next, I tackled another bug related to the Tavus integration. The in-app frames for the AISEO board and Calendly booking had been blocked, which made it impossible for users to access these tools seamlessly. This was an oversight from my previous implementation where I hadn't accounted for cross-frame interactions. I spent another focused hour diagnosing the issue, reviewing the interaction configuration for the radio-stage tools within #SPACE. Once I understood how the frames were being blocked, I was able to adjust the settings to unblock them, ensuring a smoother experience for users. This fix was particularly gratifying as it significantly enhances the usability of the Tavus integration, which is a key component of our overall offering.
Looking back on the day, I'm reminded of the unique challenges of building this platform solo. While I have Claude Code in VS Code as my assistant, there are moments when I wish I had a dedicated team to bounce ideas off or to help expedite the debugging process. Yet, there’s something incredibly rewarding about identifying and solving these issues myself. Each fix not only contributes to the product but also helps me grow as a developer. It's a constant learning curve, and I find satisfaction in the incremental improvements I make each day.
As I reflect on my progress, I can see how these enhancements feed into the larger vision of creating a one-man-show company. The goal has always been to build something that can scale and serve users effectively. Today's work on local fetch and the Tavus fixes is another step in that direction. I’m focused on ensuring that every feature I ship resonates with the user experience and adds real value.
In addition to the technical challenges, I’m also keeping a close eye on the broader implications of my work for SEO and visibility. The local fetch feature, for instance, is not just a backend improvement; it directly impacts our ability to rank for hyper-local keywords. I’m continually refining my approach to ensure that I’m not just building a functional product, but one that also stands out in search engine results. This dual focus on functionality and visibility is key to my long-term strategy.
Looking ahead, I plan to continue refining these features and addressing any other bugs that come up. The journey of building this platform is fraught with challenges, but each solved issue brings me closer to my goal. I’m excited to see how these enhancements will improve the user experience and contribute to the overall success of the platform. As I keep pushing forward, I’m confident that with each commit, I’m laying the groundwork for a billion-dollar valuation in the future.