Meta's Muse Preferred for Planning a Trip to Europe

On a trip to France and Spain, Jay Eskenazi tested several AI agents to organize his stay. According to him, Muse from Meta stood out for its practical features and useful responsiveness on-site, while Grok Bot, Instinct, and a self-hosted solution posed limitations or friction.

Instinct deemed opaque and the DIY path abandoned after costs and time

Jay Eskenazi reports that he gained access to Instinct via a waiting list the day before. He feels he doesn't see exactly what the tool does: Instinct provided him with ideas and an itinerary, but placed them in a Google Doc. He also mentions receiving personalized recommendations without recalling having provided the necessary information and questioned their origin. Prior to this, he recounts trying the self-hosted option with OpenClaw in February. The setup took him a lot of time, with the help of a developer friend for hours, followed by AWS bills he didn't understand and uncertainty about the token costs. He explains that he eventually stopped. In his view, the level of service he seeks resembles that of a very experienced travel agent from 30 years ago, a standard that would have made vacations very expensive, reminding him of his time at Expedia.

Between Claude and Grok Bot, a verdict overturned after arrival

To prepare his European itinerary, Jay Eskenazi initially worked with Claude and Grok Bot. He says he uses Grok a lot and made an update to try Grok Bot, with which he developed a sophisticated itinerary, iterated several times and satisfactory before departure. However, once on-site, he believes that this plan was not as good as expected. He asserts that Muse turned out to be the most useful for his planning, noting that he thinks he accessed it two days before leaving. He adds that he is currently traveling in France and Spain with the help of Muse.

Maps, export to Google Maps, and ticket tracking: Muse's contributions

On the eve of a day that didn't suit him, he asked Muse for an afternoon program and received numerous proposals, including new options. Unlike other tools that merely list places, he indicates that Muse positions the stops on a map with points and allows one-click sending to Google Maps, an experience he says he had over the weekend, with no equivalent in the other tools tested. For routes involving multiple museums with varying hours or free entry on certain days, he believes the AI structures everything as if one had a thousand hours, and this grouping remains effective even when applied to a thousand decision points. He also mentions a one-hour Uber ride to Versailles during which he received an organized schedule for Thursday indicating morning openings, free entries, and midday and evening plans. Unsure if ChatGPT would provide this information, he specifies that he gave Muse access to his email: the tool knew his departure and, through an ideas tab, suggested planning the opening night before his arrival. He adds that Muse checked the schedules and availability of tickets and tracked the diminishing options as he hesitated over his plans.

Method, context of the stay, and use of personal notes

Jay Eskenazi explains that he accumulated detailed notes, lists, and addresses of cafes, which he wanted to see organized in a logical sequence rather than a generic seven-day itinerary. He lives in Los Angeles and indicates that he has been working in user experience for 19 years. At the time of speaking, he is in Paris and is leaving for Barcelona the next day for a three-week trip. He says that last year, during his first long solo stay, the preparation was very stressful and research-heavy, and that this year his notes are lengthy again. Regarding Barcelona, he states that he chose to arrive in the evening on the date of an annual festival, while noting that the main activities were not published the week before on a Spanish website that did not yet display the complete program.