Wired Tests 6 Major AI Agents: Who Wins?
Original: I Tested Every Major AI Agent for a Day. Here’s the Best One
Why This Matters
Consumer AI agents are moving from niche demos to mainstream smartphone apps with real pricing tiers.
Wired reporter Reece Rogers spent a day running six always-on AI agents — Meta's Muse, OpenAI's Dots, Anthropic's Claude Duet, Google's Gemini Spark, xAI's Grok Bot, and Instinct — through identical real-world tasks to rank their reliability and usability.
After a year of frustrating demos and broken promises, AI agents have improved enough to warrant a fresh look — at least according to Wired's Reece Rogers, who stress-tested six of the biggest players in a single day. The lineup: Meta's Muse (free, chart-topping since its September launch), OpenAI's Dots ($100/month, baked into ChatGPT), Instinct (free, invite-only, runs via SMS), Google's Gemini Spark ($20/month, still in beta), xAI's Grok Bot ($30/month), and Anthropic's Claude Duet ($20/month, released as a waitlisted research preview the night before testing). Each agent was installed on Rogers' iPhone before bed, then put through five identical tasks the next morning. Agents work by controlling a virtual browser to complete jobs like scanning email, tracking flight deals, and flagging shopping opportunities — proactively pinging users rather than waiting to be asked. OpenClaw, the scrappy open-source agent credited with sparking much of the current wave, was excluded for being too complex for average users. The full rankings and task results are behind Wired's paywall, but the framing is clear: these tools have had their glow-up. They still make mistakes, but browser control and task follow-through are meaningfully better than a year ago.