engblogs

A technical exploration of multi-stakeholder AI policy and governance, advocating transparency, rigorous safety standards, public accountability, and broad access over corporate or single-actor influence.

Practical, shared playbook for trustworthy third-party evaluations of frontier models, detailing harness design, claim framing (capabilities, safeguards, comparisons), and validity reporting to inform emerging safety standards.

OpenAI launches Rosalind Biodefense to operationalize frontier AI for biodefense, enabling trusted developers and public-sector partners to advance epidemiology, early warning, and countermeasure readiness to strengthen societal resilience.

Braintrust uses Codex and GPT-5.5 to turn customer feature requests into working preview branches within minutes, enabling real-time iteration, faster customer feedback, and expanded engineering experimentation.

Boston Children’s builds an enterprise AI layer that weaves AI into clinical, research, and administrative workflows—cutting costs, expanding capacity, and enabling 40+ diagnoses that were previously unresolved through automated processes and enhanced decision support.

A practical, shared playbook for designing independent, credible evaluations of frontier-model safety and capabilities, with emphasis on harness design, elicitation, and clear validity reporting.

OpenAI's Rosalind Biodefense launch and expanded GPT-Rosalind access enable frontier AI-driven biodefense capabilities that strengthen public health, preparedness, and biosecurity through trusted partnerships.

Braintrust speeds turning customer feature requests into code by using Codex (powered by GPT-5.5) to create preview branches in minutes, enabling real-time customer feedback and rapid iteration.

A practical, shared playbook for independent, harness-aware evaluations of frontier models that credibly test capabilities, safeguards, and cross-model comparisons while emphasizing validity, evidence, and standardized reporting.

Explores how Rosalind Biodefense and GPT-Rosalind enable trusted, responsible frontier AI to advance biodefense, public health, and societal resilience through high-impact defensive applications and accelerated preparedness.

An enterprise AI layer at Boston Children’s turns AI into infrastructure, enabling rapid operational gains and clinically actionable diagnoses—driving 60,000 hours saved, $7M redeployed labor, and 40+ rare-condition identifications.

Braintrust leverages Codex and GPT-5.5 to turn customer feature requests into preview branches in minutes, enabling real-time iteration and faster feedback with customers.