Vapi Hosts Fleet Week Yacht Event: Voice AI Builders Discuss GPT Live and Natural Interaction
On October 8, Vapi chartered a yacht in San Francisco Bay to host a Fleet Week event, with more than 30 voice AI builders in attendance; Deepgram, Kouper Health, and Coval led fireside chats and technical workshops. The most talked-about topic on board was GPT Live: skipping traditional transcription/synthesis can reduce latency and make conversational turn-taking smoother, but it comes at the cost of fine-grained control over key terms and scripts. The focus of discussion has shifted from "whether voice AI can converse" to "how to make conversation more natural," specifically including backchanneling, perceived latency, and measuring naturalness.
Body
Vapicon 2026 is comingFleet Week with Vapi - RecapVapi Editorial Team • Oct 20, 20262 min readShare
On October 8, over 30 voice AI builders joined the Vapi team on a chartered yacht in San Francisco Bay to connect and watch the Blue Angels practice for Fleet Week.
Our goal was simple: voice AI builders get invited to plenty of standard networking events and rarely go. A regular office workshop has to compete with so much noise. Watching jets from a boat on the Bay, however, who could say no?
The agenda featured fireside chats and talks hosted by Deepgram, Kouper Health, and Coval, followed by a hands-on technical workshop.
GPT Live was easily one of the most discussed topics on board.
We spoke with a principal engineer at a major consumer marketplace who already deployed a live agent with it, an enterprise PM comparing it to their custom pipeline, and several others evaluating costs and ROI.
Key takeaway: skipping traditional transcription/synthesis cuts latency and creates super fluid turns, but you trade off granular control over key terms and scripting. Most teams are still figuring out which specific use cases benefit most from speech-to-speech vs classic pipelines.
The question is no longer "Can voice AI converse?" - the focus has completely shifted to "How do we make it feel natural?"
Top UX themes discussed included: backchanneling (listening noises like "mhm" or "got it" to prevent dead air), perceived vs measured latency (an 800ms pause feels long unless filled naturally), and measuring naturalness (quantifying conversational quality beyond simple latency metrics).
We also saw huge interest in embedding voice natively inside web/mobile apps (like CRMs) rather than treating it as a standalone phone channel.
Even everyday use cases like appointment scheduling carry hidden complexity when callers change their mind mid-sentence.
Enterprise Trends: Platform vs. In-House
Many enterprise teams present had already built internal voice stacks. These led to incredible conversations, as teams with custom builds know exactly where maintenance overhead gets heavy and where things break.
Why choose a platform over direct model APIs? It comes down to reliability and future-proofing. Managing multi-provider failover and staying updated across STT, LLM, and TTS models is a massive maintenance headache to own long-term.
Pro Tip for Sales & GTM Teams: The most immediate entry point isn't always greenfield innovation; it's active contact center migrations with existing budgets and a strong appetite for modernization.
Key Takeaways & What's Next
The boat format worked remarkably well. Spending four hours together in a shared space sparked organic discussions on topics like interruption handling that rarely happen at standard conferences.
A big thank you to our co-hosts Deepgram and Coval, as well as everyone who joined us.
Table of Contents