engblogs

Parameter Golf demonstrates how an open, AI-assisted machine learning challenge can accelerate research by constraining artifacts, leveraging coding agents, and surfacing new optimization, quantization, and test-time strategies that reshape evaluation and talent discovery.

NVIDIA engineers leverage Codex with GPT-5.5 to accelerate end-to-end ML research and ship production-grade systems, automating experiments, surfacing bugs, and iterating from concept to execution.

Explains how finance teams can use ChatGPT Work to transform inputs from dashboards, forecasts, and prior reviews into review-ready, shareable assets for monthly business reviews, variance analysis, and planning without coding.

Parameter Golf shows how AI-assisted research powered by coding agents and a tight, open benchmark can accelerate experimentation, surface diverse optimization ideas, and reshape evaluation and review in ML.

Learn how finance teams can turn inputs from dashboards, forecasts, and prior reports into review-ready, code-free deliverables with ChatGPT Work for monthly reviews, variance analysis, and planning.

An engineer-facing guide to boosting AI agent reliability via a concrete testing harness that measures and optimizes skill invocation across Codex and Claude.

A practical guide to building a repeatable test harness for AI agent skills, measuring skill invocation reliability with Codex and Claude, and applying actionable techniques to boost agent performance across repositories.

Explores Progressions, a modular iPad tool for quick chord-progression input, export to iReal Pro and MIDI, and expandable sound through a protocol, adapters, and Patchwork-style tool integration.

How a latency-obsessed inference platform (Modular Cloud) enables real-time, multi-modal experiences like Inkwell by streaming text and diffusion outputs with TTFT overlap, prefetching, and a unified API.

Co-Scientist introduces a multi-agent AI partner to accelerate research, highlighting extensive collaboration and the broad network of contributors behind the project.

OpenAI unleashes DeployCo, a dedicated enterprise AI deployment unit that embeds Forward Deployed Engineers to design, build, and scale production AI systems across organizations.

A concise look at how the OpenAI Campus Network unites student clubs worldwide to enable hands-on AI learning, student-led events and research, and early access to tools on AI-native campuses.