"Successful AI agents will need to have access to all your files and folders, all your data and preferences, and even other accounts and systems that you have access to as well. Are we prepared to hand it over? The answer is yes."
Seems like a company like Apple will have a huge advantage here compared to some of the other giants where we won't want to open up everything to hit e.g. Google's servers. Thoughts?
That makes a lot of sense, Apple will have a lot of your data and access to services, so will web browsers, even IDEs are in a good spot since they can actually create and manage files and folders. Things like MCP will open up more channels to other information. They will be the "API moment" of agents
"Once we cross the line where the level of trust is satisfactory, then these barriers will more and more likely get unlocked."
Reminds me of people not trusting online payments pre-2000. It will seem unthinkable in 15 years that we didn't trust agents, but we'll also forget how quickly the models progressed and how new agents felt in 2025
"At what point will we trust these agents to issue refunds or make code changes? The benefits outweigh the risks here..."
I could see companies taking a human-in-loop approach if an AI agent flags a refund above a certain dollar amount or if there are signs of abnormal end-user behavior, kind of like fraud alerts for credit card payments today
"Successful AI agents will need to have access to all your files and folders, all your data and preferences, and even other accounts and systems that you have access to as well. Are we prepared to hand it over? The answer is yes."
Seems like a company like Apple will have a huge advantage here compared to some of the other giants where we won't want to open up everything to hit e.g. Google's servers. Thoughts?
That makes a lot of sense, Apple will have a lot of your data and access to services, so will web browsers, even IDEs are in a good spot since they can actually create and manage files and folders. Things like MCP will open up more channels to other information. They will be the "API moment" of agents
"Once we cross the line where the level of trust is satisfactory, then these barriers will more and more likely get unlocked."
Reminds me of people not trusting online payments pre-2000. It will seem unthinkable in 15 years that we didn't trust agents, but we'll also forget how quickly the models progressed and how new agents felt in 2025
"At what point will we trust these agents to issue refunds or make code changes? The benefits outweigh the risks here..."
I could see companies taking a human-in-loop approach if an AI agent flags a refund above a certain dollar amount or if there are signs of abnormal end-user behavior, kind of like fraud alerts for credit card payments today