Tops GPTs 
OpenAI’s Economic Research finds Codex adoption expanded from engineers to every department by mid-2026, with non-developers growing fastest. The report says users increasingly ask Codex to complete long-horizon tasks and that Codex now accounts for most internal output tokens.
OpenAI said HP Inc. will expand use of its Frontier partnership after pilots that sped engineering, security and customer workflows. HP plans to deploy Frontier across customer and partner solutions, telemetry, employee productivity and software development.
OpenAI Economic Research published The AI Jobs Transition Framework for the EU on June 29, 2026, extending a U.S. model to European labour markets using ESCO and Eurostat data. The report maps four occupational archetypes and gives country-level shares and policy recommendations.
OpenAI reported that crashes in Rockset were traced to two separate causes: silent hardware corruption on an Azure host and an 18‑year‑old race condition in GNU libunwind. A population-level analysis of core dumps enabled the team to separate and address both issues.
OpenAI released Genebench-Pro, a benchmark package that includes ten case studies with original prompts, datasets, and evaluation materials. The tasks span somatic oncology, functional and regulatory genomics, statistical and population genetics, and single-cell analysis.
MUFG began a phased rollout of ChatGPT Enterprise at Mitsubishi UFJ Bank in 2026 after starting work with OpenAI in October 2024. The bank reports mandatory training, 1,800+ custom GPTs created in four months and a 20–30% workload drop in selected research tasks.
Australian Payments Plus (AP+) says it used ChatGPT Enterprise and Codex to reduce research and development time, speed technical investigations, and create working simulations faster. AP+ reported several internal metrics and examples of shorter investigation and prototyping cycles.
OpenAI’s audit of SWE-Bench Pro found widespread task issues and estimates roughly 30% of tasks are broken. The review identified overly strict tests, underspecified prompts, low-coverage tests and misleading prompts as main failure modes.
ChatGPT Work, introduced July 9, 2026, is an agent that can act across apps and files to complete multi-step projects. Powered by GPT‑5.6 and integrated Codex features, it runs on web, mobile and a new desktop app with Scheduled Tasks and Sites.
On July 9, 2026, OpenAI announced that GPT-5.6 will be the preferred model in Microsoft 365 Copilot, covering Word, Excel, PowerPoint, Chat and Cowork. OpenAI and Microsoft say the update aims to deliver more work per token and better cost performance for complex tasks.