GitHub's Native Stacked PRs Land in Public Preview; OpenAI Cuts GPT-5.6 Luna Pricing 80%

GitHub's long-requested stacked PR workflow goes public-preview today; OpenAI simultaneously cuts Luna 80%, Terra 20%, and adds Fast Mode for Sol.

GitHub Stacked PRs in public preview (July 30). After years of third-party tools like Graphite filling the gap, GitHub ships native stacked pull requests. Stacking lets you break a large change into an ordered series of focused PRs — each shows only its layer's diff, a stack map shows how layers relate, and merging the bottom PR lands it plus all unmerged layers above in one operation; upper PRs auto-rebase. Install the CLI extension and you're off:

gh extension install github/gh-stack
gh stack create   # from your feature branch

Works on github.com, CLI, mobile, and with GitHub Copilot agents. Merge queue support is coming.

OpenAI GPT-5.6 price cuts + Fast Mode (July 30). Effective today: GPT-5.6 Luna drops 80% ($0.20/M in, $1.20/M out), Terra drops 20% ($2/$12). Fast Mode replaces Priority Processing for Sol — up to 2.5× faster at 2× standard cost; existing priority requests auto-route to it.

Why it matters: Stacked PRs changes review ergonomics for iterative development — and with Copilot already driving multi-PR agent fan-outs, native stacking makes those reviews tractable. Luna's price cut makes OpenAI's fastest agentic model practical at scale; if you route between Claude and OpenAI in a pipeline, the cost math just shifted significantly.

Sources: GitHub Stacked PRs public preview — GitHub Changelog · About stacked pull requests — GitHub Docs · Advancing the price-performance frontier — OpenAI