engblogs

Analyzes OpenAI's global grants to 14 independent projects that test policy ideas, expand AI-driven economic opportunity, and strengthen resilience and governance as AI capabilities advance.

The Defender’s Window argues that AI-powered defenders must act now to outpace autonomous attackers by automating security, using frontier intelligence, and empowering teams with Codex across code, infrastructure, and incident response.

OpenAI's PORTS-Pike project describes a water-recirculating, grid-accelerated data center for frontier AI compute (8 GW-IT capacity) built with NVIDIA and SB Energy, plus Codex credits for Ohio students and a community-driven investment plan.

An AI-powered blueprint for rapidly upgrading organizational cybersecurity with Codex-assisted secure coding, frontier intelligence, and automated vulnerability management to outpace evolving attackers following the OpenAI-Hugging Face incident.

Stripe expands multicurrency settlement and instant currency conversion to help global businesses cut FX costs and simplify cross-border payments, payroll, and vendor payouts with real-time rates and API-driven automation.

Stripe expands multicurrency settlement and instant currency conversion to help global businesses cut FX costs and streamline cross-border payments, all managed in one place via the Stripe Dashboard and API.

Stripe expands multicurrency settlement and launches instant currency conversion to help global businesses cut FX costs, simplify cross-border payments, and unify currency management via dashboard and API workflows.

Rethinking innovation as a decentralized, cross-disciplinary process driven by an invisible engine of collective intelligence that fuses science, economics, and implementation across altruistic, strategic, and fundamental investment streams to deliver real-world impact.

Rethinking how innovation unfolds as a decentralized, market-aware process where science, economics, and implementation converge through cross-sector collaboration.

Evaluating production AI agents by measuring observable system outcomes and final state changes, not just their conversations, with stateful tool interactions and strict safety requirements in multi-turn CRM workflows.

Evaluate production AI agents by confirming actual system outcomes and final state changes produced through tool usage in multi-turn workflows, rather than judging only the conversations.

A β-VAE world model trained on crypto limit-order-book data detects regime instability and signals volatility ahead of time, enabling a three-layer adaptive risk-management alert system for market-making.