Within two points, billed as credits

The GitHub blog timestamps the Universe post at 16:09:28 UTC on October 28, 2026. The byline names Jay Parikh. The post starts from that day's Universe and names the GitHub Copilot app first. The page says the app is built for agentic software delivery and connects agents and tools across projects. The stated uses go past writing code. The page says the same surface can be used to write specs, to present design discussions, or to analyze telemetry. Those sentences describe the product. They do not carry a score.

Copilot Fusion is the passage about choosing models. The page says Fusion looks at what a task needs across reasoning, code generation, debugging, and tool use, then decides how to run the work across one model or more than one. The only numbered comparison comes next, and the page places it on the Artificial Analysis Coding Agent Index. Fusion Max scored within two points of Opus 5.5 Max, at half the cost, and with 40 percent less time per task. Within two points does not say which score is higher. Half the cost and 40 percent less time do not come with the underlying scores, and they do not come with the prices that the word half refers to.

Copilot projects are described as available starting today, and as native in the GitHub Copilot app. The page says they let agent work run across multiple repositories, in sessions that are fully managed, and in parallel. Progress, decisions, and the next step are described as living in one place. The page does not say how many parallel sessions a project can hold, and it does not say which repository hosts are included in the phrase multiple repos.

Session continuity is one cloud agent session, followed across surfaces. The page names the GitHub app, VS Code, the CLI, and the web. The example on the page runs from a phone at a standup, to VS Code at a desk, to a check on github.com at lunch, while the agent keeps working in the cloud. The same passage mentions a multi-model harness and calls it powerful, fast, and cost-efficient. It does not state a preview status or a general-availability status for that harness. Those words appear in a separate post.

Automations are workflows inside a repository. They run in the cloud. The page says they can be started by hand, started from an event, or run on a schedule. The examples of coordinated work are issue triage and code review. The new sentence says automations can learn from their outcomes and propose improvements for a person to review and approve. What gets stored as an outcome, and where the approval sits, is not spelled out.

Copilot code review Max is introduced on the same day. The page says it posted the highest F1 score on ReviewBench for high-severity issues. Lite and Balanced are described as the settings that keep everyday reviews fast. Max is the deeper look when a change needs it. The F1 number itself is not printed. The page does not attach the comparison table that would show which other review settings sit under that highest score.

Billing is the last change, and the page dates it to today. GitHub Enterprise customers can use all Copilot features without buying a separate subscription. The features named in that sentence are the GitHub Copilot app, code review, automations, and more. The charge is for the AI credits used. The path today is GitHub Sales. Self-serve is described as coming soon. Customers already on Copilot Business or Copilot Enterprise can keep the current plan or move later. The picture with this column is the Universe '26 graphic from the post.