A few weeks ago, Salesforce Research released a tool to test AI agent performance when using MCP servers. They’ve now published results of their MCP-Universe benchmark, which show that leading LLMs including ChatGPT, Google Gemini, Grok and Claude all fail more than half the time at real-life tasks such as location navigation, browser automation, and web search. Reality can be so annoying.
One of the trends we’re following is technology offerings by service vendors. Latest example is Shopify design agency Domaine, which just released an “AI Commerce Suite” that uses large language models to automate copy, imagery and video creation; real-time adaptive SEO and geo-targeting; and on-site search, recommendations, and dynamic content. That would have been impressive not so long ago; today, it’s almost commonplace.
The US Director of National Intelligence, Tulsi Gabbard, said the UK will not require Apple to provide backdoor access to encrypted data as was mandated earlier this year. Apple had subsequently withdrawn its encryption feature for new users in the U.K and disabled it for users. Apple and the UK have not commented.
Tools to help brands measure and optimize their AI search results are now too common to mention. But it’s still unusual for a CDP vendor to provide one. So let’s note that CDP Zeta Global is adding a Generative Engine Optimization solution to its AI Marketing Cloud.