Product Updates
Where to find release info and updates:
- VS Code Update Log: https://code.visualstudio.com/updates/
- Copilot Update Log: https://github.blog/changelog/label/copilot/
- Copilot CLI Update Log: - https://github.com/github/copilot-cli/releases
- Copilot SDK Update Log: - https://github.com/github/copilot-sdk/releases
GitHub Copilot Changelog - September 2026
Last reviewed: September 30, 2026. This is a dated announcement archive, not a guarantee of current availability. Check the linked product documentation for rollout, plan, policy, preview, and retirement conditions; historical launch announcements remain here for reference.
- Complete September 2026 GitHub Changelog archive
- September 30: HydraFusion in VS Code and the GitHub Copilot app
- September 29: GPT-6.1 Sol in GitHub Copilot
- September 28: Claude Sonnet 5.5 in GitHub Copilot
- September 25: Usage metrics API adds pull request review stages
- September 25: Updates to GitHub Copilot for Slack and Microsoft Teams
- September 25: GitHub Copilot weekly releases - September 21
- September 25: Enterprise managed settings in-product validator
- September 25: Agentic autofix now uses Copilot Memory
- September 24: Default enablement of Copilot features for Copilot Business and Enterprise
- September 23: More ways to request and configure Copilot code reviews
- September 23: Local sandboxing in the GitHub Copilot app
- September 22: OpenTelemetry in the GitHub Copilot app
- September 22: OpenAI’s GPT-6 Sol and GPT-6 Luna now available
- September 22: New features and improvements in Copilot for JetBrains
- September 22: Faster C++ code intelligence with whole codebase indexing
- September 22: Claude Opus 5.5 is now available in GitHub Copilot
- September 21: Grok 4.7 is now available in GitHub Copilot
- September 18: Upcoming deprecation of selected GitHub Copilot models in mid-October
- September 18: GitHub Copilot weekly releases - September 14
- September 18: Copilot code review: An improved review experience
- September 17: Copilot impact dashboard now shows feature engagement
- September 17: Agentic CLI customizations now in the usage metrics API
- September 16: Copilot budget increase requests are generally available
- September 15: GitHub Copilot suggests custom properties definitions
- September 14: Configure cost and quality in Copilot auto model selection
- September 11: Auto-resolution and analysis updates in Copilot code review
- September 11: Add VS Code Agents to Copilot usage metrics
- September 10: MAI-Code-1-Flash deprecated
- September 10: GitHub Copilot weekly releases - September 7
- September 9: Enterprise managed permissions for GitHub Copilot agent operations
- September 8: Enterprise-managed sandbox in Copilot for JetBrains
- September 4: GPT-6 Astra is generally available in GitHub Copilot
- September 4: GitHub Copilot weekly releases - August 31
- September 3: Upcoming deprecation of selected GitHub Copilot models
- September 3: Reopening Copilot Business and Enterprise signups
- September 3: Gemini 3.8 Flash is now available in GitHub Copilot
- September 2: Enterprise-managed settings support any default model
- September 2: Content exclusions generally available in Copilot app and CLI
- September 1: Set an expiration date for individual user budgets
- September 1: Copilot code review can now approve pull requests
- September 1: Claude Fable 5.1 is generally available in GitHub Copilot
Build 2026
- Build 2026 Book of News v1.1
Third party report, but generated directly from transcripts of every session – very detailed, very cool!
New Applications
- GitHub Copilot app is now available in technical preview! (GitHub Blog - May 14, 2026)
- GitHub Copilot app installer
VS Code Updates
Billing
Recent News
- Copilot budget-increase requests (September 16, 2026; GA) - Business/Enterprise usage-based billing; not available for enterprises with managed users. Owners or billing managers can approve, adjust, or deny requests.
- Auto cost and quality tiers (September 14, 2026) - efficiency, balance, and intelligence in VS Code, CLI, and Copilot app. Billing follows the selected model, not a fixed tier price; paid-plan Auto usage receives a 10% model-cost discount.
- Expiration dates for individual user budgets (September 1, 2026; GA) - time-bound Business/Enterprise overrides revert to the next applicable budget when they expire.
-
OpenTelemetry in Copilot app (September 22, 2026) - trace agent activity through enterprise-managed telemetry. Use billing reports for charges; traces are not invoices.
- Optimizing GitHub Copilot Cost in the Usage-Based Billing Era (MS Blog - July 7, 2026)
- Set AI credit session limits in Copilot CLI and SDK (GitHub Blog - July 1, 2026)
- Per-user AI credit budgets available for cost centers (GitHub Blog - June 30, 2026)
- Cost centers now support enterprise teams (GitHub Blog - June 25, 2026)
-
Included usage caps for cost centers: Helps admins limit how much of the shared enterprise AI Credit pool a cost center can draw for improved governance. (GitHub Blog - July 2, 2026)
Current guidance (September 30, 2026): Included usage controls limit a cost center’s draw from the shared AI credit pool. Administrators can view the cap and consumption on the cost center home page. See Budgets for usage-based billing and budget setup instructions.
TIP: Included usage controls are different from cost center budgets, which cap metered charges after the shared pool is exhausted. User-level budgets cap each user’s consumption across both phases. License additions or upgrades increase the included cap immediately; removals or downgrades decrease it at the next billing cycle. Moving a licensed member between two cost centers with included usage controls recalculates both caps at the next billing cycle.
Usage Based Billing is here!
- Announcement: GitHub Copilot usage-based billing is live! (GitHub Blog - June 1, 2026)
- GitHub Copilot Billing Docs
- Learn about Usage-Based Billing: GitHub Billing Platform Controls (GitHub in-depth course)
Getting Started Checklist
- Usage Based Billing Worksheet GitHub Learn site/checklist (created by GH employee TJ Corrigan in May 2026)
Original Announcement - April 2026
- Announcement: GitHub Copilot is moving to usage-based billing (Apr 27, 2026)
- GitHub Community Post with UBB FAQ (Apr 17, 2026)
- April Usage Data available to help estimate new usage-based billing impact (GitHub Blog - May 12, 2026)
Resources
- Usage-based billing for organizations and enterprises (GitHub Docs Apr 27, 2026)
- Preparing your organization for usage-based billing (GitHub Docs)
- Understanding Copilot budgeting (GitHub Support; redirects to sign-in, account access may be required). Public alternative: Copilot budget controls.
- GitHub Billing Platform Controls (GitHub Docs)
- Managing AI credits (GitHub Well-Architected Support Article)
- gh-ulb — A GitHub CLI Extension for setting GHCP User-Level Budgets (GH CLI Add-in)
What is my Cost?
- How can I see how much this costs??? (More links about tokens and optimization)
Model Pricing Snapshot (as of September 30, 2026)
Prices are in USD per 1 million tokens. This snapshot shows default-tier pricing; models with long-context tiers have higher rates above their token thresholds.
Availability is separate from pricing. A listed rate does not guarantee access in your plan, client, or organization. Consult supported models and the retirement notices below; a pricing page may retain entries that are no longer selectable.
- October 2 retirements: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7. Migration notice.
- October 19 retirements: Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, and Grok 4.5. Migration notice.
| Provider | Model | Category | Input | Cached Input | Cache Write | Output |
|---|---|---|---|---|---|---|
| OpenAI | GPT-5 mini | Lightweight | $0.25 | $0.025 | — | $2.00 |
| OpenAI | GPT-5.3-Codex | Powerful | $1.75 | $0.175 | — | $14.00 |
| OpenAI | GPT-5.4 | Versatile | $2.50 | $0.25 | — | $15.00 |
| OpenAI | GPT-5.4 mini | Lightweight | $0.75 | $0.075 | — | $4.50 |
| OpenAI | GPT-5.4 nano | Lightweight | $0.20 | $0.02 | — | $1.25 |
| OpenAI | GPT-5.5 | Powerful | $5.00 | $0.50 | — | $30.00 |
| OpenAI | GPT-5.6 Luna | Lightweight | $0.20 | $0.02 | $0.25 | $1.20 |
| OpenAI | GPT-5.6 Sol | Powerful | $4.00 | $0.40 | $5.00 | $20.00 |
| OpenAI | GPT-5.6 Terra | Versatile | $2.00 | $0.20 | $2.50 | $12.00 |
| OpenAI | GPT-6 Astra | Powerful | $10.00 | $1.00 | $12.50 | $50.00 |
| OpenAI | GPT-6 Luna | Lightweight | $0.10 | $0.01 | $0.125 | $0.50 |
| OpenAI | GPT-6 Sol | Powerful | $2.00 | $0.20 | $2.50 | $10.00 |
| OpenAI | GPT-6.1 Sol | Powerful | $2.00 | $0.10 | $2.50 | $10.00 |
| Anthropic | Claude Haiku 4.5 | Versatile | $1.00 | $0.10 | $1.25 | $5.00 |
| Anthropic | Claude Sonnet 4 | Versatile | $3.00 | $0.30 | $3.75 | $15.00 |
| Anthropic | Claude Sonnet 4.6 | Versatile | $3.00 | $0.30 | $3.75 | $15.00 |
| Anthropic | Claude Opus 4.7 | Powerful | $5.00 | $0.50 | $6.25 | $25.00 |
| Anthropic | Claude Opus 4.8 | Powerful | $5.00 | $0.50 | $6.25 | $25.00 |
| Anthropic | Claude Opus 4.8 (fast mode preview) | Powerful | $10.00 | $1.00 | $12.50 | $50.00 |
| Anthropic | Claude Opus 5 | Powerful | $5.00 | $0.50 | $6.25 | $25.00 |
| Anthropic | Claude Opus 5.5 | Powerful | $4.00 | $0.20 | $5.00 | $20.00 |
| Anthropic | Claude Sonnet 5 | Versatile | $2.00 | $0.20 | $2.50 | $10.00 |
| Anthropic | Claude Sonnet 5.5 | Versatile | $2.00 | $0.20 | $2.50 | $10.00 |
| Anthropic | Claude Fable 5 | Powerful | $10.00 | $1.00 | $12.50 | $50.00 |
| Anthropic | Claude Fable 5.1 | Powerful | $10.00 | $0.25 | $12.50 | $50.00 |
| Gemini 3.5 Flash | Lightweight | $1.50 | $0.15 | — | $9.00 | |
| Gemini 3.6 Flash | Versatile | $0.75 | $0.075 | — | $3.75 | |
| Gemini 3.7 Flash | Versatile | $0.75 | $0.075 | — | $3.75 | |
| Gemini 3.8 Flash | Versatile | $0.75 | $0.075 | — | $3.75 | |
| Microsoft | MAI-Code-1.1-Flash | Lightweight | $0.20 | $0.02 | — | $1.20 |
| xAI | Grok 4.5 | Versatile | $2.00 | $0.50 | — | $6.00 |
| xAI | Grok 4.6 | Versatile | $2.00 | $0.50 | — | $6.00 |
| xAI | Grok 4.7 | Versatile | $2.00 | $0.50 | — | $6.00 |
| Moonshot AI | Kimi K2.7 Code | Versatile | $0.95 | $0.19 | — | $4.00 |
| Moonshot AI | Kimi K3 | Powerful | $3.00 | $0.30 | — | $15.00 |
Source: Models and pricing for GitHub Copilot (GitHub Docs)
- Cache Write legend: A dash means no separate cache-write rate is listed in the source; it does not mean the interaction is free.
- Promotional pricing: GitHub lists Gemini 3.6, 3.7, and 3.8 Flash at $0.75 input, $0.075 cached input, and $3.75 output per million tokens through December 31, 2026. The earlier model retirement dates above still apply; the promotion does not extend availability.
- AI credits: 1 AI credit equals $0.01 USD. An interaction’s cost depends on its actual input, output, cached-input, and applicable cache-write tokens. See the source for long-context thresholds and rates.
Other Links and Articles
- GitHub Copilot individual plans (i.e. – NOT Enterprise or Business Plans)
- GitHub’s AI billing shift signals the end of free enterprise tools era (InfoWorld - May 12, 2026)
- Announcement: GHCP Code Reviews will start consuming GH Actions minutes (Apr 27, 2026)
Agents
- Using Agents in VS Code
- Agent Debug Logs
- AGENTS.md The generic general guide to creating agents.
- Enabling 3rd Party Agents in VS Code How can you enable 3rd party agents like Claude Code in VS Code?
- AGENTS.md outperforms skills in our agent evals - Vercel (Jan 2026) They got different results when putting data in AGENTS.md vs putting in just skills…?
- Governing agents in GitHub Enterprise (GitHub Docs)
Agent Controls - September 2026
- Enterprise-managed permissions for agent operations (September 9) - deny/ask/allow rules for Copilot app, CLI, and VS Code Agent Host sessions; user approvals cannot weaken managed restrictions.
- Enterprise settings validator (September 25) - find invalid policy configurations and team mappings.
- Content exclusions for Copilot app and CLI (September 2) - exclude configured files from agent context.
- Copilot app local sandboxing (September 23; public preview) - project-level filesystem, network, and credential boundaries; app and CLI configurations are separate.
Blogs and Articles
YouTube Videos
- Multi-agent workflows in VS Code (VS Code Channel) Demo on how you launch and run multiple agents in the background
- Agent Skills in VS Code (James Montemagno) How to build your own agent skill and use it in VS Code
- The complete guide to Agent Skills (Burke Holland) A comprehensive walkthrough of creating and using agent skills
- After This Video, You’ll Actually Understand Agent Orchestration (Burke Holland)
Agentic Frameworks
Note: these are NOT frameworks in the sense of an SDK like Microsoft’s Agent Framework. These are just some great examples of people and teams creating cool frameworks for using agents in the process
Squad
AI agent teams for any project - built by MS engineer Brady Gaster
- Squad: AI agent teams for any project (Docs) (Documentation)
- Squad: AI agent teams for any project (GitHub) (Source)
- Squad: DeepWiki Docs
- How Squad runs coordinated AI agents inside your repository (GitHub Blog)
- .Net Rocks! Podcast: Building Software using Squad with Brady Gaster (March 19, 2026)
HVE-Core
A refined collection of instructions, prompts, agents, and skills to start your project off right, or upgrade your existing projects to get the most out of all Copilots. (built by a Microsoft team)
Squad + HVE-Core ??!!!
An installable package that assembles HVE components into a bundle and then ships them with a Squad Coordinator! (created by an MS engineer)
Git-Ape
An intelligent Azure deployment agent system for GitHub Copilot. Provides guided, safe, and validated Azure resource deployments using ARM templates, with built-in security analysis, cost estimation, and CI/CD pipeline integration. (built by a Microsoft team)
- Git-Ape (Documentation)
- Install Git-Ape (VS Code Marketplace Plugin)
- Git-Ape Source (GitHub Repo)
Apex
Transform Azure platform engineering for IT Pros. Powered by GitHub Copilot and specialized AI agents, APEX converts requirements into architecture diagrams, governance-aligned designs, and deploy-ready Bicep or Terraform templates — with built-in pricing, compliance, and automation. (a fairly complex repo built by an MS engineer)
- Agentic Platform Engineering eXperience for Azure (Starting page for docs and repo)
Agentic Workflow
Official Guides
- GitHub Agentic Workflows (official repository) - define AI-powered repository automation in Markdown and run it through GitHub Actions.
- Agentic Workflows quickstart (official documentation) - prerequisites, authentication, and adding a first workflow. Review the workflow’s permissions and schedule before enabling automation.
Videos and Sessions
- Introducing GitHub Agentic Workflows | intent-driven repository automation (YouTube short intro)
- GitHub Agentic Workflows: Automation That Actually Reads the Room (Build 2026 Session)
Articles and Blogs
Articles and videos below reflect their publication dates. Third-party analysis is not official product guidance, and some publishers may require sign-in or a subscription.
Recurring Interesting Posts
How it works
- The Coding Harness Behind GitHub Copilot in VS Code (VS Code Blog - May 2026)
- Components of A Coding Agent Explaination of how Coding Agents work (3rd party blog)
Best Practices
- How VS Code Team Builds with AI (VS Code Blog - March 2026)
- The AI Native Mind (Written and maintained by Daniel Meppiel - a Software Global Black Belt at Microsoft)
- Read: The Agentic SDLC Handbook (by Daniel)
- The Agent Package Manager (by Daniel)
Worst Practices
- Uber CEO, COO says AI spending not yet tied to measurable improvements MSN Summary - May 28, 2026
TL/DR: Uber burned through all of it’s 2026 token budget in 4 months, and the COO says he cannot link 70% AI-generated code or 95% engineer adoption to measurable consumer product improvements.
- Claude just delete my entire production environment! (and who’s fault is that…??!) (April 2026)
- Anthropic Warns That “Reckless” Claude Mythos Escaped a Sandbox Environment During Testing (Futurism Blog: April 2026)
- Someone left Claude Code running overnight, and it cost $6,000 (Make Use Of Blog - May 2026)
Articles - Industry Trends
- Tokenomics is the new headcount Jared Spataro Blog (Microsoft CMO of AI@Work)
- Blog: The state of global AI diffusion in 2026 (MS Blog - May 2026)
- Global AI Diffusion Report 2026 (Q1 2026 Trends and Insights)
Articles - Our Careers
- How tech workers are feeling in 2026: a workforce splitting in two (July 2026 blog summarizing a survey of 5,920 tech professionals)
TL/DR: One half feels amplified by AI—more capable, more confident, more excited than they’ve been in their entire career. The other half feels shaken by it—less sure of their value and whether there’s still a place for them. The workforce is bifurcating into two realities.
- Redefining the Software Engineering Profession for AI (Mark Russinovich & Scott Hanselman)
- From Implementers to Solution Architects (Medium - Feb 2026 - good insights on how our career is changing)
- Where I’m feeling: behind… (April 2026 Blog)
- Coding is solved? Software is not. (Arcplane.ai Blog - May 2026)
- The AI Vampire (Steve Yegge Blog - Feb 2026)
Articles - Cost Optimization
- Managing AI Coding Costs at Scale (databricks AI blog Aug 2026)
Articles - Techniques
- How Drasi used GitHub Copilot to find documentation bugs (Microsoft Open Source Blog)
- Agents are power tools (GitHub vNext Blog - May 2026)
- AGENTS.md outperforms skills in our agent evals (Vercel Jan 2026) - They got different results when putting data in AGENTS.md vs putting in just skills… ?
Interesting Podcasts
- The AI Vampire with Gas Town’s Steve Yegge (Scott Hanselman’s podcast - Feb 2026)
- How AI Changes Development with Rob Conery (DotNetRocks Podcast April 2026)
- Leading Teams in the Time of AI with Andrew Murphy (DotNetRocks Podcast February 2026)
- GitHub’s plan for Agents (Latent Space Podcast with Kyle Daigle, COO/CMO GitHub - June 2, 2026)
- The Role of AI in Testing with Debbie O’Brien Great conversation on testing with AI and Playwright (DotNetRocks Podcast July 2026)
- The Story of VS Code Documentary on YouTube (~98 minutes - released Sept 4, 2026)
- Software Fundamentals Matter More Than Ever Matt Pocock (YouTube — August 2026)
Azure DevOps
Recent News
- Azure DevOps and GitHub: Journeying into the AI Era (MS Blog - June 2, 2026)
- Azure DevOps meets GitHub, the path to AI powered SDLC (Video - MS Build 2026 Breakout)
Presented by Dave Burniseon and Dan Hellem - the two most knowledgable people I know on this topic…!
See also - GitHub Repo for this session
New Feature Announcements
- Copilot Code Reviews for Azure Repos Preview (Public Preview - August 26 2026)
- Copilot Code Reviews for Azure Repos Announcement (Technical Preview - June 2026)
Checked September 30, 2026: The August announcement opened the public preview. The current setup documentation still labels the feature a limited preview and warns that capabilities roll out in stages. Confirm availability and enablement in your organization rather than relying on the original rollout estimate.
Usage is billed through the Azure subscription linked to the Azure DevOps organization and appears in Azure Cost Management. Consult the setup documentation for current billing details; do not assume your GitHub Copilot plan’s included credits cover Azure Repos reviews.
- Copilot Code Reviews for Azure Repos Announcement (Technical Preview - June 2026)
- Copilot Autofix for GitHub Advanced Security for Azure DevOps - Blog (Technical Preview Feature - June 2026)
- Copilot Autofix for Code Scanning - Documentation (Technical Preview Feature - June 2026)
Migrating to GitHub
- Enterprise Live Migrations (preview; checked September 30, 2026)
ELM provides CLI and portal experiences for continuous synchronization and controlled cutover from Azure DevOps Services to GitHub Enterprise Cloud with data residency. Azure DevOps Server repositories must first migrate to Azure DevOps Services. ELM support in the Azure DevOps Remote MCP Server is also in preview. See the linked documentation for prerequisites and current limits.
- How Microsoft is migrating repositories to GitHub (MS Dev Blog - June 2026)
- Migrating from Azure DevOps to GitHub (GitHub Docs)
- Azure DevOps to GitHub Migration Playbook: Unlocking Agentic DevOps (Microsoft Dev Blogs)
- Azure DevOps to GitHub Enterprise Migration Guide (GitHub Well-Architected Guide)
- Migrating Azure DevOps to GitHub Enterprise: The ROI Case (Adam the Automator - 3rd Party Blog)
MCP
- Set up the remote Azure DevOps MCP Server (preview) (MS Learn)
- Azure DevOps Remote MCP Server (preview) (Azure DevOps Blog - March 2026)
- Install Git DevOps Assistant (MCP Powered) (VS Marketplace)
Learning Path
- Optimize DevOps with AI agents on Azure (Microsoft Learning Path - 7 Modules - Updated June 2026)
- Introduction to agentic DevOps using Microsoft tools on Azure
- Manage Azure Boards by using GitHub Copilot integration
- Manage Azure DevOps by using MCP Server
- Optimize Azure reliability by using Azure SRE Agent
- Use GitHub Copilot plan mode for cloud operations tasks
- Use GitHub Copilot for cloud operations
- Transition from traditional ops to agentic DevOps workflows
Boards (non-AzDO)
Jira
- GitHub Copilot coding agent for Jira public preview announcement (March 5, 2026)
- Integrate Jira with GitHub | Atlassian Support (general development-activity integration; distinct from Copilot cloud agent setup)
- Integrating Copilot cloud agent with Jira (current setup guide; checked September 30, 2026)
Prerequisites: A paid GitHub Copilot plan, a Jira Cloud account with Jira enabled for AI, and Rovo activated for your organization. Both the Jira and GitHub applications must be installed and authorized; users need write access to the target GitHub repository to trigger the agent. Check your Atlassian plan and entitlement rather than assuming a separate Rovo purchase is always required.
Cheat Sheets
What feature should I use?
Select the relevant client tab in GitHub Docs; commands and customization support vary by client. For terminal commands, use the Copilot CLI command reference.
Code Indexes
C++ Symbol Indexing in Copilot CLI
- Whole-codebase indexing for C++ (GitHub Changelog - September 22, 2026) - reuses a persistent symbol index for definitions, references, and implementations. Enabled by default in the Microsoft C++ Language Server;
/lsp logsis a Copilot CLI command.
This language-server symbol index is distinct from the semantic workspace indexes described below.
Repository and Workspace Indexing
- Indexing repositories for GitHub Copilot - GitHub Docs
- How Copilot understands your workspace (VS Code documentation) - search tools, semantic index sources, and workspace exclusions
Checked September 30, 2026: Copilot automatically manages workspace indexing, but local and remote index sources still exist. GitHub repositories use GitHub indexing, Azure DevOps repositories require Microsoft sign-in for their managed indexes, and other workspaces can use automatically built indexes subject to account policies. Remote indexing is not supported for GitHub Enterprise Server. Check the Copilot status dashboard in the VS Code Status Bar; use Build Codebase semantic index in the Command Palette to request a build or rebuild. Text search and grep can still work without a semantic index.
Copilot CLI
Official Guides
- Copilot CLI quickstart (GitHub Docs) - installation, authentication, and first interactive or scripted prompts.
- Copilot CLI command reference (GitHub Docs) - CLI options and slash commands; use
/helpto check your installed version. - Managing context in Copilot CLI (GitHub Docs) -
/context,/compact, and long-session behavior.
Courses and Articles
- GitHub Copilot CLI for Beginners (GitHub Course- April 2026)
- Install GHCP CLI
- CLI - Interactive vs. Non-Interactive Mode (GitHub Blog Apr 30, 2026)
- GitHub Copilot CLI for Beginners (GitHub official course/repo)
- GitHub Copilot CLI Extensions: The Complete (unofficial) Guide(blog Hector Flores - Microsoft CSA)
Videos
- YouTube: You need to try the GitHub Copilot CLI right now! (Burke Holland)
- YouTube: Getting Started with GitHub Copilot CLI and Work IQ (John Savill - March 2026)
- YouTube: How to extend Copilot CLI with custom UI (Steve Sandersen - Apr 23, 2026)
Short Tips
- YouTube Short (2 min): Copilot CLI - use TWO models with subagent calls (Burke Holland)
Copilot SDK
Official Guides
- GitHub Copilot SDK (official repository) - supported languages, installation, examples, and API references.
- Copilot SDK getting started (official guide) - build your first application; check the prerequisites for your chosen language.
Videos
- YouTube: The GitHub Copilot SDK is out!
- YouTube: GitHub Copilot SDK Tutorial: Build AI Apps with Live Coding
- YouTube: GitHub Copilot SDK demo: Creating “Flight School” Chris Reddington demonstrates “Flight School,” a custom Next.js application built to personalize his learning journey using the GitHub Copilot SDK. See how he leverages agentic workflows to generate daily coding challenges based on his GitHub profile, evaluate solutions against test cases, and automatically export projects to new repositories.
General Repos
- Awesome GitHub Copilot (UI Tools)
- Awesome GitHub Copilot (Source)
- florinpop17/app-ideas: A Collection of application ideas which can be used to improve your coding skills
- github/spec-kit Toolkit to help you get started with Spec-Driven Development
Labs and Workshops
- Copilot CLI for Beginners (Great beginner repo)
- Mona Mayhem: Build an app that compares GH contributions (Workshop)
- Mona Mayhem: Build an app that compares GH contributions (repo)
GitHub Roadmap
- GitHub Public Roadmap (official repository)
- GitHub Project Board (GitHub Projects; sign-in or project access may be required)
Roadmap items are plans, not guaranteed release dates; use the official changelog to confirm shipped features.
GitHub Copilot Initial Setup
Instructions for setting up a full GitHub EMU account, creating Enterprise Teams and onboarding Copilot only users. This set-up will not require GitHub Enterprise seats.
Step 1: Create GitHub Enterprise Managed User (EMU) Trial
- Start here
- About Enterprise Managed Users
- Getting started with Enterprise Managed Users
- You can configure authentication either by using OIDC set up or SAML. We recommend using OIDC which includes support for CAP.
- Configuring OIDC for Enterprise Managed Users
- Configure SSO for your GitHub Tenant (Video)
Step 2: Create Enterprise Teams
Step 3: Add an Azure Subscription
Step 4: Assigning GitHub Copilot Business licenses via Enterprise Teams.
- Assign GitHub Copilot Licenses (Video)
Other Reference Links:
- Tutorial Videos: Setting up GitHub Copilot for your Company | All things Azure
Great short video series covering the entire process
- Configure a GitHub enterprise with Enterprise Managed Users for SAML Single sign-on with Microsoft Entra ID (Microsoft Learn)
- Configure GitHub Enterprise Managed User for automatic user provisioning with Microsoft Entra ID (Microsoft Learn)
Enterprise Management
September 2026 Governance Updates
- Enterprise-managed agent permissions (September 9) - centrally deny, require approval for, or allow shell commands, file operations, and network domains. Applies to Copilot app, CLI, and VS Code sessions using Agent Host.
- Managed settings validator (September 25) - validate managed settings, team mappings, and team settings files in enterprise AI controls before relying on their enforcement.
- Content exclusions in Copilot app and CLI (September 2) - review enterprise, organization, and repository exclusion policies during onboarding. See current content-exclusion support and limitations.
- Enterprise-managed sandbox policies in JetBrains (September 8; public preview) - check preview availability and managed restrictions for your IDE.
- Local sandboxing in Copilot app (September 23; public preview) - off by default, configured per project, and separate from CLI sandbox settings.
Earlier Management Resources
- Deploy managed Copilot settings via MDM in VS Code and CLI (GitHub Change Log - July 8, 2026)
Note: Some of the settings let you disable Bypass Permissions and set allowed marketplaces
Tips:
Two new features (July 2026) for Copilot Enterprise admins to apply even more control where needed:
- managed-settings.json This lets the enterprise centrally define governance and policy for Copilot across the org — managed as code, in one place.
- Enterprises can now default to Auto model selection Admins can set model to auto in the enterprise managed-settings.json to make Copilot’s auto model selection the default for new conversations.
GitHub Tokenomics Course
Command scope: The slash-command examples below are for Copilot CLI, not every Copilot client. Check the CLI reference or /help for your installed version; IDE and app controls may differ.
September 2026 Follow-up Resources
- Auto cost and quality tiers (September 14) - choose efficiency, balance, or intelligence in supported VS Code, CLI, and Copilot app versions.
- OpenTelemetry in Copilot app (September 22) - enterprise-managed agent tracing; prompt/response capture is off by default.
- Temporary budget expiration (September 1) and budget-increase requests (September 16) - Business/Enterprise governance; requests are not supported for enterprises with managed users.
Five levers, ordered by leverage
- Context hygiene: Clear between tasks, compact proactively, resume when you can.
- Prompt discipline: Plan before coding, reference specific files and lines, not directories.
- Model selection: Match model tier to task complexity each session. Choose per phase not mid-conversation.
- Scope & tool control: Tight working directory, content exclusion, minimal custom instructions.
- Measurement: CLI
/usage,/context, and telemetry make waste visible.
Ten Things You Can Do Today
- Choose the right model for the task: use Auto mode, escalate only when needed
- Give clear, focused guidance in your prompts
- Work in phases - fresh window each time: Research -> Plan -> Implement
- Add deterministic guardrails: tests, linters, security scans
- Keep a concise, human-written copilot-instructions.md
- Start a new session per task: /clear or /new between unrelated work
- Reference files with @path/to/file, not whole directories
- End every session with /usage: know what it cost
- Open with /plan for anything beyond a one-line change
- Start small – start now: run /usage after your next three sessions. You’ll know where to begin.
If you remember only one line from today - remember this
Write as little context as required, and as much as necessary
Links Shared in Presentation
- AI Model Comparison (GitHub Docs)
- Models and Pricing for GitHub Copilot (GitHub Docs)
- Changing the AI model for GitHub Copilot Inline Suggestions (GitHub Docs)
- Project HydraFusion: Frontier quality via multi-model orchestration (The GitHub Blog - September 4, 2026)
- Satya Nadella’s Post on HydraFusion (LinkedIn September 5, 2026)
- Context Rot: Why AI Gets Worse the Longer You Chat (Blog)
- Improving Token Efficiency in GitHub Agentic Workflows (GitHub Blog)
- Improving Token Efficiency for GitHub Copilot in VS Code (VS Blog)
- Token Economy: Optimizing GitHub Copilot Chat & Agents under Usage-Based Billing (GitHub Repo Community Playbook)
- AI Engineer Coach (VS Code Extension)
- Monitor AI Coding Agents with Grafana (Microsoft Learn)
- Improve Agent Quality and Token Optimization (GitHub Support; redirects to sign-in, account access may be required). Public alternative: Copilot CLI context management.
- Optimizing GitHub Copilot Cost in the Usage-Based Billing Era (Azure Architecture Blog)
- AI@Work: Tokenomics is the new headcount (Jared Spataro Blog - Microsoft CMO of AI at Work)
Presentation Slides
- GitHub Tokenomics Presentation (September 8, 2026)
Hooks
- Using hooks with GitHub Copilot CLI (GitHub Docs)
Examples
- Agent Hooks: Deterministic Control for Agent Workflows (3rd party blog/repo with some great examples)
Labs and Workshops
Microsoft Learning Paths
- Get started with AI-assisted development MS Learning Path
6 modules and ~8 hours of independent study course materials focused on learning how to use GitHub Copilot in VS Code
- Optimize DevOps with AI agents on Azure Microsoft Learning Path
7 modules focused on integrating AI-powered agents into your Azure DevOps workflows (checked September 30, 2026).
Labs
- Great beginner lab repo (workshop using GHCP CLI)
- Build a cool app from scratch using the GHCP CLI (from Copilot Days)
- Learning Hub | Awesome GitHub Copilot
- Day in the Life Copilot Lab Awesome hands-on lab from Cody Carlson
My Super Simple Demos
- Build an Agent Demo (Renato Camara’s demo repo)
- Build a Skill Demo (Renato Camara’s demo repo)
- Build an MCP Server Demo (Renato Camara’s demo repo)
- Build a Copilot CLI Extension Demo (Renato Camara’s demo repo)
- Build a C# program that calls a Foundry Model with Managed Identity using the GHCP SDK (Renato Camara’s demo repo)
Languages
C++ in Copilot CLI
- Faster C++ code intelligence with whole-codebase indexing (GitHub Changelog - September 22, 2026) - persistent project-wide symbol indexing in the Microsoft C++ Language Server. In Copilot CLI, use
/lsp logsto check indexing progress; the initial build can take extra time and memory. - C++ code intelligence for GitHub Copilot CLI (public preview April 2026)
Learning Paths
- microsoft/Mastering-GitHub-Copilot-for-Paired-Programming: A multi-module course
- GitHub Copilot CLI for Beginners (GitHub official course)
- microsoft/mcp-for-beginners: This open-source curriculum introduces the fundamentals of Model Context Protocol (MCP)
- Terminal-First AI Pair Programming with GitHub Copilot CLI (LinkedIn Learning - February 2026; sign-in and course access may be required)
- Introduction to agent-first development in VS Code (VS Code Learn Site)
- Intro to Agent Customizations (VS Code Learn Site)
Podcasts
- VS Code Insiders Podcast (Monthly podcast on new features)
MCP
- MCP Documentation
- MCP Inspector Tool - Source
- Using MCP Servers in Visual Studio
- MCP developer guide (VS Code)
- OWASP MCP Top 10 Security Guidance for Azure (Microsoft Docs)
- OWASP MCP Top 10 (OWASP)
Labs
- MCP Hands-On Lab
Cody’s Lab on using MCP servers
- Simple MCP Server Lab
Renato Camara’s Lab on building a very simple MCP server using C#
- MCP Security Summit Workshop
Great hands-on workshop that walks you through how to secure MCP Servers in Azure using APIM, defensive postures, monitoring, etc.
MCP Management
- Managing MCP usage in your company (GitHub Docs)
MCP Registries
Other MCP Servers
- Azure DevOps MCP Server (Microsoft Docs)
Memories
- Copilot Memory has more controls for deletion, scope, and the Copilot CLI
- Managing and curating Copilot Memory (GitHub Docs)
Memory Scope and Controls (checked September 30, 2026)
- User-level preferences are visible only to you and can be used across your repositories. Review or delete them in your personal Copilot Memory settings.
- Repository-level facts are shared with repository contributors. Repository owners can review, delete, or disable them under Settings > Copilot > Memory.
- Disabling repository memory stops repository facts from being stored or read; it does not delete existing facts or disable user-level preferences.
- In Copilot CLI, use
/memory on,/memory off, and/memory showto manage the persisted setting. Organization and enterprise policies can affect availability.
The May 26 announcement above documents these controls as released functionality, not a future rollout.
Sessions
- How to query your Copilot CLI session history (Evan Boyle - GHCP CLI Team)
Model Links
- AI model comparison - when to use what models (GitHub Docs)
- Hosting of models for GitHub Copilot - provider-specific hosting, retention exceptions, and plan-dependent training policies (GitHub Docs)
- Supported AI models in GitHub Copilot (GitHub Docs)
- Models and pricing for GitHub Copilot (GitHub Docs)
- Base and long-term support (LTS) models (GitHub docs)
- Microsoft Foundry Models - Lifecycle and Support Policy
- Microsoft Foundry Model Retirement Schedule
Model Lifecycle (checked September 30, 2026)
- Retired September 10: MAI-Code-1-Flash; GitHub recommends MAI-Code-1.1-Flash instead.
- Scheduled for October 2: Gemini 3.5 Flash, Gemini 3.6 Flash, Kimi K2.7 Code, and Claude Opus 4.7. GitHub suggests Gemini 3.8 Flash for both Gemini models, Kimi K3 for Kimi K2.7 Code, and Claude Opus 5 for Claude Opus 4.7.
- Scheduled for October 19: Gemini 3.7 Flash, GPT-5.5, GPT-5.4, GPT-5.4 mini, GPT-5 mini, and Grok 4.5. Check the announcement for migration choices and your organization’s model policies before updating workflows.
Historical Announcements
- Building a hill-climbing machine: Launching seven new MAI models (Microsoft Build 2026 Announcement)
- Introducing MAI-Code-1-Flash (historical; retired from Copilot September 10, 2026)
- Use your own language model key (BYOK) in VS Code (MS Blog - June 18, 2026)
- Enterprise bring your own key (BYOK) for GitHub Copilot is now in public preview - GitHub Changelog (Nov 20, 2025)
Anthropic Announcement History
- June 9, 2026: Claude Mythos/Fable is released
- June 12, 2026: Claude Mythos/Fable is blocked
These June announcements are historical, not a current Copilot availability statement. As of September 30, GitHub lists Claude Fable 5 and 5.1 with access and data-handling conditions. Review supported models and hosting and retention requirements before enabling them.
Interesting Articles
- Intelligence Per Dollar (3rd party blog - Tomasz Tunguz)
- Independent analysis of AI (Great independent model comparison website)
Quick Model Usage Summary (checked September 30, 2026)
These are starting points, not a ranking or a guarantee of availability. They draw on GitHub’s model comparison and the September release announcements. Check your plan, client, rollout, and administrator policies. Auto is a useful starting point where available.
| Task | Models to evaluate |
|---|---|
| General-purpose coding and agent tasks | GPT-5.6 Terra, GPT-5.3-Codex |
| Well-scoped features and bug fixes | Claude Sonnet 5.5 |
| Fast help with simple or repetitive tasks | GPT-5.6 Luna, Claude Haiku 4.5 |
| Deep reasoning and debugging | GPT-5.6 Sol, Claude Sonnet 4.6 |
| Agentic coding and terminal workflows | GPT-6.1 Sol |
| Working with visuals | Claude Sonnet 4.6; confirm image input support in your client |
Model Selection
Last reviewed: September 30, 2026. This is a task-based starting guide, not a fixed inventory of deployed models or an official ranking. Availability depends on your Copilot plan, client, rollout, and organization policies.
Start With the Current Catalog
- Supported models and client availability
- GitHub’s model comparison by task
- Token pricing and long-context rates
- Auto model selection
Use Auto where available when you do not have a specific model requirement. For explicit selections, evaluate quality, latency, and total task cost on representative work instead of assigning one model permanently to every agent or skill.
Models to Evaluate by Task
These examples combine GitHub’s task guidance with the September release announcements. They deliberately avoid models already retired or scheduled for retirement in October.
| Task | Starting candidates | What to evaluate |
|---|---|---|
| Routine coding, documentation, and everyday agent work | GPT-5.6 Terra, GPT-5.3-Codex | Correctness and clarity on your codebase |
| Well-scoped features and bug fixes | Claude Sonnet 5.5 | Completion quality, tool calls, and retries |
| Small edits, explanations, and repetitive tasks | GPT-5.6 Luna, Claude Haiku 4.5 | Cost and responsiveness without losing accuracy |
| Multi-step reasoning, debugging, and architecture | GPT-5.6 Sol, Claude Sonnet 4.6 | Cross-file correctness and evidence-backed conclusions |
| Agentic coding and terminal workflows | GPT-6.1 Sol | Successful completion, token use, and verification |
| Visual input | Claude Sonnet 4.6 | Image support in the actual Copilot client you use |
- Claude Sonnet 5.5 release and plan availability: the announcement lists Pro, Pro+, Max, Business, and Enterprise, with gradual rollout.
- GPT-6.1 Sol release and plan availability: the announcement lists Pro+, Max, Business, and Enterprise, with gradual rollout.
Separate Model, Reasoning, and Context Settings
- Model: Select a supported model from the client picker or use the identifier documented for that client’s configuration format. Display names are not necessarily valid configuration identifiers.
- Reasoning effort: Where supported, this is a separate control. Labels such as “High” or “XHigh” are not independent model releases and should not be appended to an identifier unless that client explicitly documents it.
- Context capacity: Long-context support depends on both model and client. A “1M” capability does not mean every session receives that context window. Check availability, thresholds, and any higher token rates before enabling it.
- Agent and skill configuration: Do not assume every client supports the same YAML fields or per-skill overrides. A skill does not inherently select a model; follow the configuration documentation for the host that invokes it.
Retirement and Migration Checklist
| Date | Affected models | GitHub’s suggested alternatives |
|---|---|---|
| September 10, 2026 (retired) | MAI-Code-1-Flash | MAI-Code-1.1-Flash |
| October 2, 2026 (scheduled) | Gemini 3.5 Flash, Gemini 3.6 Flash | Gemini 3.8 Flash |
| October 2, 2026 (scheduled) | Kimi K2.7 Code | Kimi K3 |
| October 2, 2026 (scheduled) | Claude Opus 4.7 | Claude Opus 5 |
| October 19, 2026 (scheduled) | Gemini 3.7 Flash | Gemini 3.8 Flash |
| October 19, 2026 (scheduled) | GPT-5.5, GPT-5.4 | GPT-5.6 Sol |
| October 19, 2026 (scheduled) | GPT-5.4 mini, GPT-5 mini | GPT-5.6 Luna |
| October 19, 2026 (scheduled) | Grok 4.5 | Grok 4.6 |
Before migrating, confirm the replacement is enabled and supported in your client, review its pricing and data-handling terms, update any explicit configuration, and rerun your task evaluations.
Fallback and Cost Guidance
- Use the client’s documented fallback behavior or explicitly select another currently supported, policy-approved model. An example list is not an automatically configured fallback chain.
- Do not downgrade silently when required capabilities, permissions, or privacy constraints cannot be preserved. Resolve the constraint or report the limitation.
- Compare total cost per successful task, including retries, tool interactions, input/output tokens, caching, and long-context pricing. No model is universally the cheapest for every task.
- Treat generated-output reuse separately from provider prompt caching; revalidate reused outputs when their inputs or underlying facts change.
- Review base and long-term support models and provider data-handling conditions before establishing an organization-wide policy.
Modernization Skill
- GitHub Copilot modernization documentation
- Closing the AI-readiness gap with agentic modernization (MS Blog)
- VS Code Extension - GitHub Copilot modernization for .NET
- modernize-dotnet/VS-CHANGELOG.md at main · dotnet/modernize-dotnet
- GitHub Copilot modernization for C++ overview (MS Learn)
- dotnet/skills: Repository for skills to assist AI coding agents with .NET and C#
Current Guidance (checked September 30, 2026)
- Modernization agent overview: the Modernize CLI documents an assess, plan, and execute workflow, reusable skills, multi-repository assessment, and batch operations.
- Modernization agent quickstart: installation instructions and prerequisites.
Check each capability’s documentation for access and preview terms rather than treating earlier roadmap announcements as a guarantee of general availability.
Security and Privacy
- Trust Center - github.com
- Secret scanning with GitHub MCP Server is now GA (GitHub Changelog May 5, 2026)
- Microsoft AI Compliance, Security and Privacy Resources (MS Service Trust Portal)
Data Handling (checked September 30, 2026)
- Model hosting and data handling: retention and training terms depend on the model, provider, feature, and Copilot plan. Zero data retention is not a blanket promise for all Copilot data.
- GitHub states that it does not use Copilot Business or Enterprise customer data to train AI models. Individual subscribers may have interaction data used for training under applicable settings and can opt out through individual Copilot policies. Provider training commitments and GitHub’s own use of data are distinct.
- Claude Fable 5 and 5.1 retain prompts and outputs by default for Anthropic safety classifiers. Eligible enterprises can request a time-bound zero-data-retention exemption through the end of 2026; it requires approval, configuration, separate model enablement, and compliance with usage restrictions. Consult the hosting documentation and your GitHub account team. Some Anthropic preview features also fall outside the standard ZDR agreement.
Sandboxes
- Local sandboxing in Copilot app (September 23, 2026; public preview) - off by default; limits filesystem, network, and credential access for local sessions. It does not apply to cloud or remote-host sessions.
- Enterprise-managed sandbox policies in JetBrains (September 8, 2026; public preview) - managed restrictions take precedence over user settings.
- Quick Intro to Windows Sandbox (John Savill - YouTube)
Enterprise Guardrails - September 2026
- Enterprise-managed permissions for agent operations (September 9) - policy-enforced deny/ask/allow rules for commands, file operations, and network domains.
- Managed settings validator (September 25) - detects malformed JSON and invalid policy configurations.
- Content exclusions in Copilot app and CLI (September 2). Review client support and limitations; content exclusions are not a replacement for sandboxing or operating-system access controls.
Skills
- Creating agent skills for GitHub Copilot
- Context-Driven Development: Agent Skills for Microsoft Foundry and Azure | All things Azure
- Announcing the Azure Skills Plugin | All things Azure (new March 9, 2026)
- GitHub Copilot now supports Agent Skills - GitHub Changelog (December 18, 2025)
Resources
- Awesome Copilot Skills (GitHub community collection)
- Anthropic Skills (Anthropic repository)
- Matt Pocock’s Skills (home of the “grill-me” skill)
Learning Paths
- Interactive “Clippy Teaches Agent Skills”
Funny (and educational!) web app featuring Clippy - vibed by Addy Osmani - Directory, Cloud AI - Google
- Agent Skills Overview Site
Tokens
September 2026 Cost and Observability Updates
- Auto cost/quality tiers (September 14) - efficiency, balance, and intelligence in VS Code, CLI, and Copilot app; charges depend on the model selected, not a flat tier price.
- OpenTelemetry in Copilot app (September 22) - enterprise-managed export of agent traces. Prompt and response content is excluded by default; review content-capture settings before enabling it.
- Individual budget expiration (September 1) and budget-increase requests (September 16) - Business/Enterprise controls; increase requests exclude enterprises with managed users.
- Copilot CLI command reference -
/usageand/contextare CLI commands, not universal commands across Copilot clients.
Token Costs
- Models and model pricing for GitHub Copilot (GitHub Docs; checked September 30, 2026)
- See your GitHub Copilot CLI tokens and estimated spend at a glance
Add-in for GHCP CLI to view cost live (repo developed by Microsoft CSA Gabriel Mercuri) - How to Size AI Credit Budgets per User
Custom Usage Dashboards
Export your usage via the GHCP API (must be a Billing admin…), then import that data into these custom PBIs and create your own dashboards.
- Copilot Credit Usage & Chargebacks - PBI Reports (MS - GitHub Repo)
- Open-source analytics tools for Microsoft Copilot and AI adoption - PBI Reports (MS - GitHub Repo)
Token Efficiency
- The Economics of Agent Optimization: Four ways to lower the cost (Azure Blog - August 2026)
- Improve agent quality and token optimization (GitHub Support; redirects to sign-in, account access may be required). Public alternative: Copilot CLI context management.
- Improving token efficiency for GitHub Copilot in VS Code (MS Blog - June 2026)
- GitHub Copilot Token Optimization Workshop (GH Recurring Webinar)
Note: You can also simply watch this previously recorded session on YouTube
- Token Economy: Optimizing GitHub Copilot Chat & Agents under Usage-Based Billing
(Repo full of tips by a MS CSA - May 2026) - How to Monitor AI coding agents with Grafana (MS Learn)
- Improving token efficiency in GitHub Agentic Workflows (GitHub Blog May 7, 2026)
This article describes token optimizations for GitHub Agentic Workflows, including replacing some MCP interactions with CLI calls and normalizing consumption by model cost. Tool-definition overhead depends on the client and its tool-loading behavior; not every MCP setup sends every tool definition on every request.
Pricing example (September 30, 2026): At the published default rates, Claude Opus 5.5 costs $4 per million uncached input tokens and $20 per million output tokens; Claude Haiku 4.5 costs $1 and $5 respectively. For equal uncached input and output counts, that is 4 times the cost, not a universal “Opus versus Haiku” multiplier. Cached-input rates differ by 2 times ($0.20 vs $0.10), and cache-write rates by 4 times ($5 vs $1.25). Actual task cost also depends on token counts, caching, and retries. See current GitHub pricing.
- What is Prompt Caching? Optimize LLM Latency with AI Transformers (IBM YouTube Channel - Feb 2026)
- Solution Optimization for GitHub Copilot Workshop
Ask your CSAM to schedule this workshop if you have Microsoft Unified support!
TokenMaximizing -> TokenMinimizing
- Tokenmaxxing: What to know about the AI status game (March 2026)
- Tokenmaxxing: Silicon Valley’s Dumbest New Status Symbol (March 2026)
- The tokenmaxxing era is over. Now companies are ‘tokenminimizing’ (June 2026)
- How AT&T Handled 27 Billion Daily Tokens, Cut Costs by 90%, and Redefined AI Governance Some great guidelines
Interesting Ideas…
- CLI proxy that reduces LLM token consumption by 60-90% on common dev commands
3rd Party GitHub Repo / Plug-in by rtk-ai- (Note: I haven’t verified or tested this yet, but it looks interesting!!!)
- Why use many token when few do trick - talk like caveman - cut 65%
3rd Party GitHub Repo / Plug-in by juliusBrussee- (Note: Caveman use few words - shrink tokens)