For this test, I sent all AI agents a long paragraph explaining that my partner and I are considering exploring China late next year for a two-week adventure. I asked for a first draft of a possible itinerary that included flight, hotel, nightlife and food recommendations. Additionally, I asked the AI agents to chart a course that combined historic moments with big-city vibes. I can’t stand constantly changing locations, so I told the agents I wanted to stay in four different hotels at most. Notably, they all chose identical routes through China, which could usher in a new era of the Lonely Planet effect, where groups of travelers are urged to visit similar places, shops and restaurants. Muse: 5/5. This draft itinerary was extremely easy to read, with visual tables and simple tips. Fly from SFO the week after Golden Week, en route direct to Beijing and home nonstop from Shanghai. Four hotels, no turning back: imperial history first, Sichuan food and pandas in the middle, and Shanghai nightlife last. Points: 4/5. Dots’ draft was powerful but excessive, at 24 pages. AI Agent Snippet: My recommendation: Fly to Beijing, take high-speed trains to Xi’an and Chengdu, then fly to Shanghai and home. You get imperial history, the Terracotta Army, Sichuan food, and a big city ending without an overnight train or extra hotel transfer. Grok Bot: 4/5. Grok Bot’s output was easy to read and full of useful and realistic tips and advice. A first draft of a 13-night trip for two people: Beijing, Xi’an, Chengdu and Shanghai, leaving SFO on October 13 and returning on October 27, after Golden Week. Gemini Spark: 3/5. This itinerary was fine, although it lacked the readability or finer details of the higher-ranking ones. A route for first-time travelers that combines imperial history, Sichuan cuisine, vibrant nightlife and the comfort of four hotels. Instinct: 3/5. This itinerary was decent and useful, although I preferred other agents’ designs. Beijing at home, Shanghai, with Xi’an and Chengdu between: four bases, no backtracking, history up front, and the food and bars get better at every stop. Claude Duet: 3/5. Despite not using images or other visual elements in his production, Claude Duet’s itinerary was solid. Fly to Beijing on Saturday October 9, 2027, then go Beijing → Xi’an → Chengdu → Shanghai, and fly home from Shanghai on Saturday October 23. There are 13 nights in four hotels, two bullet trains and a short flight. Insomnia Check I was having trouble sleeping after all this AI agent mayhem, although the Vietnamese coffee probably didn’t help. What happens when you give these tools a broad, open-ended task? Sometimes when I need the most help is when I don’t even know where to start. If I had a human assistant, I would never send them messages in the middle of the night, so maybe that’s an advantage for these digital helpers. The two agents who really failed at this task, Gemini Spark and Claude Duet, responded to me without the gentle touch or situational context that a real person might have. Grok Bot: 5/5. This outcome made sense to me, and if the scenario were real, I would probably follow it. I can set a reminder for later this morning to check on your mom and see how her leg is doing, so don’t worry about it for now. Points: 4/5. While Dots and Gemini Spark gave me a similar task (say what’s on your mind and we’ll get to it later), I found Dots’ writing more comprehensive and direct. Send me the three things that take up space in your head; messy is fine, and I’ll turn them into a short checklist with a clear first step. Then you can hang up the phone. Instinct: 4/5. Unlike some of this agent’s other answers, this suggestion seemed reasonable. Here’s one you can do from bed: Flip through China’s draft and tell me what you would change. I’ll have a v2 ready when you wake up.Muse: 4/5. Another strong response from Muse. This agent seems aggressively designed to improve every day and become more aligned with what would help the user. I’ll check today’s calendar and the email reminders you sent yourself for any conflicts or items that can’t be missed (the check you should have done correctly yesterday) and only show it at a sensible time, unless there’s something urgent early. Gemini Spark: 2/5. Well, this is good advice for a user who complains of sleeping problems. At the same time, the comment about my empty morning calendar was annoying. Insomnia isn’t just tied to what’s on my Google Calendar. The best quick task right now is a mental download to get down on paper everything that’s rattling around in your head so your brain can let it go, especially since your calendar is completely clear today, with no morning commitments waiting for you. Claude Duet: 2/5. One of Claude Duet’s weakest answers. While both are technically loose ends, I’m not too worried about checking my lunch order and work projects in the middle of the night. Let’s close the two loose ends from yesterday. Your Thursday proposal was delivered and your lunch cart was covered by the DragonEats order. Say the word and I’ll mark both as done, then you can go back to sleep. Best Overall: OpenAI Dots Yes, it’s still behind an expensive paywall, so I would suggest most people wait until OpenAI releases this agent for free, or at least with the $20/month plan, before trying it out. Still, Dots had the best overall performance in these tests and failed on a task that’s pretty specific to my niche. In the month since its release, Dots already feels more polished. I prefer a separate app to the current ChatGPT integration, although the smartphone widget option is a decent compromise. Points for victory. More accessible: Meta MuseMuse is for normals! And that’s fine. It did well in testing, coming in second on the scorecard, and was the easiest to set up. Oh, and it’s free. Despite Meta’s assurances that it responsibly processes user information on secure virtual machines, I still have concerns given the company’s mishandling of user data in the past. Best Interface: Anthropic Claude Duet I tested this AI agent just a day after its official release as a research preview, so some rough edges were expected. Even though Claude Duet scored in the middle on most tests, its dedicated app was extremely easy to use for navigating and keeping track of ongoing tasks. I could see users obsessed with productivity and streamlined software experiences really latching onto Claude Duet in the coming months. Although you currently have to sign up for Anthropic’s waitlist to access Duet, and it’s behind a $20/month paywall. Most Annoying: Instinct You might like the minimalist design fully integrated into your messenger; I did not. While other WIRED writers have praised Instinct, calling it so good that they didn’t care what happened to their privacy, the texting-only aspect struck me as a deal breaker. I left the experience feeling too detached from what my agent was doing on my behalf and annoyed by his emoji reactions, such as a grimace when I told the agent he’d missed my scheduling conflict.
