---
type: "post"
id: "github-intelligence-puts-your-code-context-and-history-771f01"
slug: "github-intelligence-puts-your-code-context-and-history-771f01"
title: "GitHub intelligence puts your code, context, and history to work"
description: "GitHub has made Copilot Memory generally available on paid Copilot plans and put Copilot dreaming into public preview."
retrieval_nugget: "GitHub has made Copilot Memory generally available on paid Copilot plans and put Copilot dreaming into public preview. Together with cross-repository code search, the release turns prior agent work, repository conventions, and accepted review feedback into context that can persist, be proposed for change, and be governed by teams."
published_at: "2026-10-30"
updated_at: "2026-10-30"
record_date: "2026-10-30"
date_kind: "published_at"
topics: ["github","code-review"]
entities: ["GitHub","GitHub Copilot","Copilot Memory","Copilot dreaming","GitHub Copilot CLI","GitHub code search"]
related_patterns: ["company-memory-needs-write-loops","harness-architecture-outlives-model-choice","verification-bandwidth-is-the-scarce-resource"]
novelty_label: "new_fact"
what_changed: "The announcement adds that Copilot Memory is generally available on paid plans, alongside broader GitHub intelligence capabilities."
prior_refs: ["memory:59869bb7-86f3-877c-a9f8-a5dae38f9a29"]
source_url: "https://github.blog/changelog/2026-10-28-github-intelligence-puts-your-code-context-and-history-to-work"
source_urls: ["https://github.blog/changelog/2026-10-28-github-intelligence-puts-your-code-context-and-history-to-work"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:github-intelligence-puts-your-code-context-and-history-771f01"
status: "published"
visuals: []
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/github-intelligence-puts-your-code-context-and-history-771f01/","markdown":"https://newruntime.com/posts/github-intelligence-puts-your-code-context-and-history-771f01.md","json":"https://newruntime.com/posts/github-intelligence-puts-your-code-context-and-history-771f01.json"}
---

# GitHub intelligence puts your code, context, and history to work

## Retrieval answer

GitHub has made Copilot Memory generally available on paid Copilot plans and put Copilot dreaming into public preview. Together with cross-repository code search, the release turns prior agent work, repository conventions, and accepted review feedback into context that can persist, be proposed for change, and be governed by teams.

## Summary
GitHub has made Copilot Memory generally available on paid Copilot plans and put Copilot dreaming into public preview. Together with cross-repository code search, the release turns prior agent work, repository conventions, and accepted review feedback into context that can persist, be proposed for change, and be governed by teams.

## What happened
GitHub has announced GitHub intelligence, which it describes as a context and learning layer sourced from the graph of GitHub data. Its purpose is to help Copilot agents retain how a developer or team works, keep that guidance current, and locate code and development history relevant to a task. The release groups the capability into three parts: Memory, dreaming, and code search.

Copilot Memory is now generally available on paid Copilot plans. It can retain personal preferences and repository conventions across sessions and Copilot clients, including the Copilot app, Copilot CLI, VS Code, the Copilot cloud agent, Copilot code review, Slack, Teams, and GitHub Mobile. A personal preference could be an instruction to begin pull-request descriptions with a TL;DR. A repository memory could record the relevant test command or the way a team expects Copilot to check its work. Users can interact with memories in the Copilot app, CLI, and VS Code; personal memories can also be viewed or deleted in Copilot settings.

GitHub says that Copilot sessions autonomously generate and curate memories from interactions. It reports that, among public-preview Copilot CLI sessions it reviewed in which the agent used memory, 93% showed better adherence to developer preferences and 42% showed more efficient workflows attributed to memories included in session context. Those are GitHub-reported observations from preview sessions, rather than an independently described benchmark.

Copilot dreaming is now in public preview. GitHub describes it as reflection on past work intended to improve future agentic sessions. Three types are available: personal memory curation, which updates memories from a user's agent conversations; repository customization, which manages repository-scoped memories and proposes updates to skills and custom-instruction files; and code-review customization, which turns recurring accepted review feedback into guidance for Copilot code review. Dreaming can add, update, or remove personal and repository memories. Changes to checked-in skills and instruction files are proposed through pull requests for team review.

Dreaming is opt-in. Individual subscribers enable personal curation in Copilot settings and repository dreams in each repository's settings. Administrators enable it for organizations and enterprises through Copilot policies, with enterprise administrators able to delegate that choice to organization administrators. Repository dreams apply to repositories permitted by Copilot Memory's repository-access settings, and users must also enable personal dreaming themselves. Dream runs consume AI credits; GitHub directs users to its documentation for prerequisites, setup, and costs.

The third part is GitHub code search for Copilot. It can search related repositories and dependencies without a local clone, using exact text and symbols, regular expressions, repository, path, and language filters, or semantic search. GitHub's example is a payment-retry investigation spanning a payments service, shared library, and background workers. Through `/research`, a user can ask Copilot to inspect implementations and callers across those repositories. GitHub positions this for migration planning, distributed bug investigation, and finding organizational standards in unfamiliar codebases.

## What the archive adds
This is a new event on a known GitHub Copilot thread. Earlier coverage described [local Copilot sandboxing](https://newruntime.com/posts/local-sandboxing-for-github-copilot-is-now-generally-dc5772/) as an isolated execution context with controlled file access. The current release addresses a different boundary: how the agent accumulates and updates context across work. A same-day note on [Copilot CLI model selection and the SDK release](https://newruntime.com/posts/github-copilot-cli-1-0-105-lets-users-77ef51/) covered model and client changes, not persistent memory or dreaming.

## Trend signal
This strongly supports [Company memory needs write and correction loops](https://newruntime.com/patterns/company-memory-needs-write-loops/). GitHub is not presenting memory as a passive retrieval layer: sessions generate and curate it, dreaming can revise or remove it, and checked-in instruction changes move through pull requests. The important product question becomes whether those write and review loops produce reliable organizational guidance.

## Why it matters
Many coding-agent failures are context failures rather than raw code-generation failures: the agent uses the wrong test command, misses a local convention, applies an outdated design assumption, or repeats feedback that reviewers have already given. Persistent memory and cross-repository search aim directly at that problem. Dreaming extends the idea by treating accepted feedback and earlier conversations as candidate updates to the agent's working guidance.

For a team, this can reduce repeated prompting and the manual assembly of context for multi-repository work. But it also creates a governance surface. Memories, skills, and instruction files can influence future changes, so teams need clear ownership over which guidance is automatically curated, which changes require pull-request review, and which repositories an agent may search.

## What happens next
- **Teams will formalize review ownership for agent-maintained instructions and skills.** (3 months, 70%). Watch for: policy controls, audit views, or repository-level approval settings.
- **Cross-repository investigation will become a standard Copilot workflow for migrations and distributed bugs.** (6 months, 65%). Watch for: more `/research` guidance and organization-wide search controls.

## Limits and uncertainty
GitHub does not establish how accurate or durable autonomously generated memories are across different repositories, languages, or team sizes. Its 93% preference-adherence and 42% workflow-efficiency figures are limited to GitHub-reviewed public-preview CLI sessions where memory was used, with no methodology, sample size, baseline, or independent evaluation in the supplied announcement. The announcement also says dream runs consume AI credits but does not provide pricing or usage limits. It does not explain how conflicting memories, stale repository guidance, or undesirable learned review habits are detected beyond the stated ability to manage memories and review proposed checked-in changes.

## Skill to build now
Create a small, reviewable repository guidance file that records only high-value facts an agent repeatedly needs: the test command, validation steps, architectural boundaries, and pull-request conventions. Assign an owner for it, define what changes require review, and separate durable team rules from a developer's personal preferences. That gives dreaming or any future memory system a clear target for proposals rather than an unbounded pile of instructions.

## Reader takeaway
GitHub is moving Copilot from session-by-session prompting toward a maintained context system. The practical opportunity is less repeated explanation and faster cross-repository investigation; the practical responsibility is to treat agent memory and evolving instructions as production artifacts that need access controls and review.

## Open question
Will team-reviewed proposals be enough to prevent repository memory from slowly accumulating stale or misleading guidance as codebases and practices change?

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Github](https://newruntime.com/topics/github.md): Explore the github topic hub.
- topic hub: [Code Review](https://newruntime.com/topics/code-review.md): Explore the code-review topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)