Signals

OpenAI's CEO says we're heading for an AI fraud crisis, and the authentication methods most companies rely on are already broken.

Ruth Porat's human-in-the-loop mandate gives product teams a concrete framework for building agentic systems that users will trust and regulators will accept.

VCs are backing companies that use AI to achieve software margins in service businesses, but research shows 40% of workers spend extra time fixing low-quality AI output that eats the projected gains.

MIT researchers tested GPT and ERNIE in English and Chinese. The finding: language choice shapes the cultural assumptions in AI responses. When prompted in English, models reflected American values. In Chinese, they shifted to Chinese values.

Andrusko's a16z analysis reveals why "thinking partner" rhetoric clashes with billable hour economics. Since AI tools will never be flawless, traceability and workflow adaptation matter more than sophistication

Anthropic's safeguards architecture demonstrates how legal frameworks transform into computational systems that process trillions of tokens while preventing harm in real-time.

Working with summer interns revealed that the next generation treats AI as another developing tool, not an existential threat.

Justice Kagan's surprise at Claude's constitutional analysis reveals an irony: while we fixate on AI hallucinations, we miss when machines reason more systematically than humans, modeling dispassionate legal analysis.