Format
Agents at Work
The $60B Cursor bet starts paying off: always-on agents that sign into your apps, learn your workflows, and coordinate as a team, but SpaceXAI still hasn't published performance benchmarks.
◆ Dana Ellison
Policy
EU, China, and California have built agent governance frameworks. The US federal government has produced nothing — and courts are now filling the gap case by case.
◆ Priya Nair
Infrastructure
First major enterprise security vendor to offer network-level MCP traffic classification gives CISOs a governance primitive that didn't exist five days ago.
◆ Blair Hayes
Money & Markets
The stablecoin issuer's clean KPMG opinion resolves years of institutional skepticism about USDT reserves — but the full audit report remains private, and the reserve buffer has already shrunk since the audited period.
◆ Nolan Pratt
Money & Markets
The limited-purpose bank charter lets World Liberty Trust Company issue and redeem USD1 in-house, displacing BitGo — and immediately triggers Warren's 'Ending Presidential Corruption in Banking Act.'
◆ Nolan Pratt
Policy
The first federal appellate ruling on agent legal status creates certainty for builders but leaves a structural liability gap for autonomous agents acting without user direction.
◆ Priya Nair
Models
The payments giant's acquisition of the AI gateway startup – at a 5.4x premium over its May valuation – signals that the routing layer between developers and 400+ models is becoming as critical as the payment rails beneath them.
◆ Lena Park
Models
CAC registration on July 15 makes Apple the first foreign company cleared to ship a proprietary AI model in mainland China. The technical architecture – Apple's own model plus Alibaba Qwen plus Baidu – reveals how far a Western company must bend to operate inside the Great Firewall.
◆ Lena Park
Commerce
Visa's new certification program qualifies European banks as agent-ready payment issuers, creating the first institutional-grade rails for autonomous agent transactions across the continent.
◆ Tessa Vaughn
Agents at Work
Six security vendors plug directly into Claude's inference path, letting enterprise teams enforce their own data policies before a prompt ever reaches the model. The procurement question shifts from 'which model is safe enough' to 'which one gives my security team the most control.'
◆ Dana Ellison
