Goldman's engineers have a new challenge: turning AI agents into firm insiders
THE SO WHAT
Goldman trying to encode "tribal knowledge" into AI agents is a clear signal that institutional memory is becoming a machine-readable asset. If you run a complex org, start cataloging the undocumented rules and edge cases you’d need to teach an agent before you trust it with real work.
READ THE SOURCE
MORE FROM THE WIRE
Applied AIIndia AI Data Center Firm Orders 9,000 Nvidia Vera Rubin Systems
An Indian provider ordering 9,000 Vera Rubin systems is another data point that AI compute is regionalizing, not just centralizing in US hyperscalers. If you’re planning large-scale training or inference in Asia, local capacity and partnerships may open up faster than expected.
Applied AIKeeping your options open: Why choice matters for UK AI sovereignty
UK AI "sovereignty" framed as choice and competition—not full stack self-sufficiency—signals a policy path that favors multi-vendor, multi-cloud architectures. If you operate in or sell into the UK, align your stack and contracts to emphasize portability and optionality.
Applied AISpirit Airlines Wants to Sell Its Data to Google. Former Flight Attendants Are Freaked Out
Selling historical employee data for AI training without clear consent is a reputational and regulatory tripwire, not just a monetization lever. If you’re exploring data deals, assume staff and customers will eventually see the fine print—go slower and over-communicate.
Anthropic tells San Francisco staff to work from home due to possible security team strike
A potential security team strike forcing remote work shows how dependent AI labs are on physical access control and on-site ops. If your core IP sits in a few offices or data centers, labor relations and contingency planning belong in your security model.