The Day We Killed the Date Picker

From the Travel Tech Podcast newsletter for episode 17, The Day We Killed the Date Picker.

I've wanted to get a solo founder building an AI-native travel product onto the show for a while. Not a team with VC backing and a GDS contract; someone who spotted a structural problem in how travel is sold, had no prior coding experience, and built a working product anyway. For those of you reading this across aviation, travel tech, and AI, I knew this conversation would land. This week, I got it.

My guest went from OECD policy analyst to expat English teacher to data economist to AI-first solopreneur, learning to code from scratch using LLMs while simultaneously earning a master's in statistics and AI in Paris. His product is live and taking users. You can find the episode on Apple Podcasts, also streaming on Spotify and YouTube.

What emerged from the conversation wasn't just a product story. It was a sharp diagnosis of whose interests the current travel stack is actually optimized for, and whether AI gives us a genuine chance to change that.

What's broken in travel search, and why it matters now

The way most people book flights was designed for a world of blue links and form fields. You pick two dates, two cities, hit search, and get a ranked list shaped by someone else's incentive structure. The platform is "free," but commissions are embedded somewhere in the price chain, visible or not. Legacy OTA models were built to monetize the transaction, which means the optimization goal was never purely the traveler's.

Now those same OTAs are being plugged into AI chat interfaces as integrations. Which means: you pay for a Claude or ChatGPT subscription, click through an OTA integration, and the OTA earns a commission you don't see. That's the internet-era intermediary problem dressed in a new interface, not a structural fix.

Christopher Olivares, founder of Elyo, built his product specifically to avoid that model. We talked about the mechanics of airline distribution (GDSs, NDC, white-label APIs, and the seller-of-record problem), why Elyo uses a freemium subscription model instead of commissions, how LLMs can translate flexible traveler intent into structured flight searches, what's wrong with the date picker, and whether AI makes the old-school human travel agent look prescient in retrospect.

Key Takeaways

• The "free" platform isn't free: OTA commissions (relatively low for flights compared to hotels, but present) are absorbed somewhere in the price chain, often by the traveler in ways that aren't visible at point of booking.
• Seller-of-record is a serious early-stage liability: Many white-label distribution APIs grant easier booking access but make the startup the seller of record for the ticket, a compliance and operational burden that most solo or early-stage founders should decline at this stage.
• LLMs are good at messy human intent: The core technical insight behind Elyo is that frontier LLMs can decompose flexible, conversational traveler goals ("cheapest city for four people to meet from different origins") into structured search queries that a computer can execute across dozens of combinations simultaneously.
• Developer portals are closing, not opening: At least one major GDS has shut down its developer portal, reducing early-stage builders' ability to experiment with live data before committing to full commercialization. A trend worth watching for anyone building in distribution.
• AI may return travel to a pre-intermediary relationship: By making advisory intelligence cheap to deliver, AI potentially enables the personalization of the human travel agent era alongside the price transparency the metasearch era created, without requiring a commission layer to pay for either.

Alex Brooker, Host of the Travel Tech Podcast, Founder of Airside Labs