Latest Signal
When AI shops for you, who owns the mistake?
Payment companies are deciding what an AI agent is allowed to spend. Those same controls may end up deciding who pays when it buys the wrong thing.
Making sense of what's changing.
Some developments make more sense when you look at them together. Future Memory follows those connections across technology, business, institutions and everyday life, and revisits what changes over time.
This week: card networks, banks and insurers all started defining what an AI agent is allowed to spend, and that turns out to decide who pays when it gets it wrong.
How one subject unfolds
AI agents and platform access, in the IndexCan a platform refuse an AI agent that its own customer has authorized?
Under reviewPerplexity moves to dismiss the amended complaint, arguing it repeats theories the Ninth Circuit already rejected and that Amazon's terms are an at-will contract.
ReportingSalesforce agrees to buy Braze, a customer-messaging platform, for about $5.66 billion ($47.50 a share in cash). Expected to close in Salesforce's fiscal 2028.
ReportingWalmart sets terms for AI shopping agents: an MCP-based Agent Gateway launching as a beta for companies and developers, agent profiles in customer accounts with approval for every purchase or a spending limit, the same prices for agents as for people, and blocking for agents that don't comply.
PrimaryA ruling on Perplexity's motion to dismiss Amazon's amended complaint
Following now
All subjectsAI agents and platform access
Who controls the customer relationship when software does the shopping?
Delegated agency
What happens to responsibility when the one deciding is not the one carrying it out?
The economics of the AI build-out
Can a real technology shift still be a poor investment for those financing it?
Change at work without layoffs
What happens to employment when organizations stop replacing people rather than announcing layoffs?
When the answer replaces the source
How is original information paid for when people no longer need to visit the source?
Surveillance and institutional power
When do extraordinary surveillance tools become ordinary, and what safeguards hold?
Who oversees the AI systems institutions depend on
Who decides whether an AI system can be trusted, and on what evidence?
Precedents we're watching
All precedentsUnder review
Can a platform refuse an AI agent that its own customer has authorized?
AI agents and platform access
Emerging
Will embedded independent evaluators become a norm for frontier AI companies?
Who oversees the AI systems institutions depend on
Emerging
Can governments bar an AI model, and products built on it, from public systems on supply-chain grounds?
Who oversees the AI systems institutions depend on
How it fits together
Signals
Short pieces on what a development means, written when there's something worth saying. Each one ends with what would change the reading.
The Index
The subjects we follow, each with a dated timeline of developments, its sources, and why it connects to other subjects.
Precedents
Developments that could set a rule for what comes next. Each has a status that changes, in public, as things move.