THE FRONTIER BRIEF

■ The Big Story

OpenAI cuts the wait to send code changes

OpenAI built Pando, a new backend for its code repository, without changing how its engineers use Git. Pando now supports development across OpenAI’s shared codebase. OpenAI says the typical wait to send code changes fell nearly fivefold in production. For a busy software team, that is time recovered from a bottleneck every change must pass through.

Greg Brockman says OpenAI deliberately kept Git so engineers would not have to change how they work. He also says OpenAI is considering providing Pando to others and has invited expressions of interest. For now, the concrete result is internal: a faster route for code changes, rather than a product other businesses can buy. The possibility of wider access is worth watching.

Sources: [OpenAI] [Greg Brockman on X]

■ Frontier Watch

■ The Street Says

The practical debate is whether faster code delivery helps teams manage growing codebases, while builders also question how much to trust striking research claims and which AI model suits a particular job.

■ Repo of the Day

■ Two Steps Ahead

Over the next 30–90 days, ask your software lead to separate time spent generating code from time spent sending, reviewing and combining it. Pando puts the second part in focus; max-sixty / worktrunk addresses the friction of working on several changes at once. If delivery is the constraint, buying more AI coding capacity will not by itself clear the queue.

OpenAI’s planned security sprint and Anthropic’s Claude Code rule point to two different vendor questions: how weaknesses get found and fixed, and which customer requests a tool will refuse. At the next review of an AI-assisted development workflow, put both questions into concrete examples drawn from your team’s work. Anthropic’s Residency also creates a near-term option for an eligible senior engineer considering a research move, with an application deadline rather than an open-ended career plan.