This weekend’s Hospitality 2030 hackathon at Rosewood Sand Hill gave me a clearer view of where AI in hospitality is heading. My biggest takeaway: AI interest is already there. Luxury hotels are exploring voice agents, guest personalization, and operational automation. The harder problem is making these systems work inside real hotel environments. A hotel runs across many systems, departments, and human touchpoints: front desk, concierge, rooms, housekeeping, spa, food and beverage, security, events, CRM, surveys, reservation systems, and daily team rituals. That fragmentation makes hospitality a uniquely challenging environment for deploying AI agents. A generic chatbot only touches the surface. The deeper opportunity is building systems that understand guest context, preserve staff knowledge, and route the right information to the right team at the right moment. In luxury hospitality, AI also has to respect the tone of service. The best experiences feel discreet, contextual, and human led. Technology should help staff create better moments while staying mostly invisible to the guest. The strongest insight I took away is that guest memory may be one of the most important missing layers. Staff hear meaningful details every day: a preference a stressful situation a favorite drink a prior issue a reason for traveling Those signals can disappear before they become action. That feels like a powerful design space for AI: human listening structured memory better service Huge shout out to @cerebral_valley and @RosewoodHotels for bringing together such a thoughtful group of builders, operators, and hospitality leaders in an exceptional setting!

At last night's AI Design Night, I asked @pablostanley (design lead at @vercel) what he thinks about Claude Design. "It's fun. Still figuring it out." My take: Claude Design will do for designers what Claude Code did for devs. The designers who figure this out in 2026 will be 10x designers. Designers, am I wrong? – at San Francisco, CA

Only 60% agreement between human annotators and Gemini on audio rubrics for voice AI evaluation. If your judge cannot tell the difference between a politely silent model and one that genuinely handled an interruption, your RL reward signal is broken from the start. @scale_AI is building hybrid audio rubric judges to close this gap. Great talk from @agxsai at a voice AI gathering in SF tonight hosted by @scale_AI

Every business defines compliance differently. A bank, a hospital, a Disney contact center each need their own guardrails. Vapi’s approach: LORA adapters, a few dozen MB, trained on 100 labeled examples per agent. Customizable, cheap, and fast enough to sit in the live speech path. Heard from @DevSeth44 from @Vapi_AI at a voice AI gathering in SF tonight! Huge shout out to @scale_AI for hostinf!

Full circle moment with @wesbos and @stolinski at @sentry event 😇 Huge fan of @syntaxfm 🙌

SF by night 🌙

Typing out code for @LeetCode questions is like touching grass 😇

Natural therapy at Lafayette Park😇Blessed to be in SF💚

Inspiring read, great insights @tankots! Where is the "- Written with @WisprFlow" tagline? Was expecting it at the end😁