TechCrunch: Testing a $6,880 AI agent for executives

TechCrunch tested Vertu’s Alphafold foldable phone, a $6,880 device marketed to executives that ships with a built-in AI agent called Hermes Agent. The hardware is a rebranded ZTE Nubia Fold with premium leather and a significant price premium. Hermes Agent is built on open-source technology and is designed to execute multi-step tasks autonomously — analyzing files, scheduling, managing communications, and escalating to a human concierge service when tasks exceed its capabilities.

Testing showed a gap between the premise and the execution. Hermes outperformed Samsung’s Gemini in some autonomous task scenarios. But it incorrectly scheduled reminders, failed to recall documents uploaded days earlier, and produced incomplete results that required manual correction. The reviewer concluded that it presents as an evolving platform rather than a finished product that justifies the price.

For product managers evaluating or building AI agents, the review surfaces issues that remain characteristic of the category in mid-2026. Context retention across sessions is still unreliable. Multi-step autonomous execution handles straightforward cases but breaks down at edge conditions. And premium pricing positioned on potential rather than proven reliability creates a market credibility risk.

These are not specific to Hermes Agent. They reflect the current ceiling of agentic AI systems asked to operate across diverse real-world tasks without domain-specific guardrails. The implication for PMs building agents is that scoping the agent to a narrow task surface with clear success criteria produces more reliable outcomes than broad autonomy designed to handle any executive need.