Key Takeaways

  • Claude Duet failed an Adidas sneaker price check because major retail storefronts repeatedly blocked its browser session, leaving it unable to sort items by price.
  • Dot skipped basic comparison logic during retail searches, refusing to check adjacent colorways when an exact red sneaker was out of stock.
  • Grok Bot ran through e-commerce stores without getting stopped by bot defenses, making it Claire Vo's default agent for live price comparisons.
  • Claude Duet outperformed every rival on a launch video for ChatPRD Product Intelligence, pairing on-brand visual pacing with voice generation from ElevenLabs.

Retail Sites Block Claude While Grok Clears the Cart

Most founders assume that a general-purpose agent with computer use can run any web task. Claire Vo tested that assumption by sending Claude Duet, Dot, Muse, and Grok Bot to find the best deal on a pair of Adidas sneakers. The test broke immediately on basic web infrastructure.

Claude Duet hit an immediate wall. “Claude totally got blocked by almost every shopping site,” Vo explained. “I don't know why. It was like consistently blocked over time.” The agent could not sort prices or bypass standard bot-detection screens. Dot did not get blocked, but its reasoning fell apart under normal consumer friction. “Dot did not do comparison,” Vo said. “It didn't look at different colors. It didn't like say like I can't find red but I can find maroon.”

Grok Bot and Muse ran circles around both. Grok moved fast, checked inventory across multiple merchants, and handled price sorting without triggering defensive site barriers. “Grock was super fast and did a good job,” Vo noted. “I have Grock do a lot of this like price comparison shopping for me.” If your workflow requires pulling prices across open retail sites, Claude Duet currently costs more time in failed requests than it saves in automation. Grok Bot is the working tool for that loop.

Claude Duet Wins Marketing Launches Through MCP Tool Calls

Claude Duet reversed the standings when Vo switched the benchmark from external web scraping to asset creation. The task: produce an end-to-end product launch video for ChatPRD Product Intelligence.

Dot treated the prompt like a static document generator. It built a series of rigid slides that looked like an outdated PowerPoint deck, completely missing the rhythm and visual hierarchy required for product marketing.

Claude Duet approached the task as an integrated production tool. “Claude absolutely crushed this task. Duet crushed this task,” Vo said. “And so what I would say is Claude for video. Claude for video, Jen. It's just it's just the best.”

The agent succeeded because it understood both brand assets and external tools. It read ChatPRD's product identity, created a clean layout without broken spacing, and reached outside its default boundaries. “Claude nailed the value prop. It used our branding. There were no layout issues,” Vo said. “It actually went and asked and used the 11 Labs MCP to come up with a voice over.” By querying the ElevenLabs Model Context Protocol (MCP) server, Claude produced a voiced asset that sounded like a human product marketer cut it.

What to Do With This

Audit the tasks you hand to autonomous agents this week and split them by operational environment. Route open-web research, live retail checks, and merchant inventory tracking directly to Grok Bot to bypass bot walls. Reserve Claude Duet for closed-loop creative tasks where you can wire it to tool servers like the ElevenLabs MCP, giving it full control over branding, script pacing, and voice generation.