Best of GitHub — June 2026

1. 1
   Article

   API Evangelist·10w

   The Architecture Behind 10,000 API Providers

   Kin Lane describes the architecture behind a catalog of 10,000 API providers built over 16 years. Each provider is represented as a GitHub repository containing an apis.yml file in APIs.json format (v0.19), which serves as a machine-readable profile declaring identity, API entries, and standardized property types like OpenAPI specs, Plans, RateLimits, and FinOps data. The catalog has indexed 10,754 OpenAPI specs and 14,123 unique tags. The APIs.json standard provides a shared vocabulary that eliminates the need for custom parsers per provider, enabling governance queries across the entire dataset. The GitHub repos feed APIs.io, a search and discovery layer. Using GitHub as storage provides version history, programmatic access via API, and community editability through pull requests.

   112

   3
2. 2
   Article

   Visual Studio Blog·11w

   Review pull requests without leaving Visual Studio

   Visual Studio 18.7 now supports full pull request review workflows directly inside the IDE, covering both GitHub and Azure DevOps. Developers can browse changed files with inline or side-by-side diffs, leave and reply to comments, resolve threads, and approve or merge PRs without switching to a browser. Reviews can be done without checking out the branch to preserve working state, or with checkout for deeper investigation using VS navigation and debugging tools. Copilot integration allows generating code fixes from review comments. Features still in progress include comment filtering and a PR activity timeline.

   94

   13
3. 3
   Article

   The New Stack·10w

   Cursor, GitLab and Zed agree GitHub is breaking. They disagree on how to rebuild it.

   GitHub is struggling under the load of AI-generated code, processing 1.4 billion commits per month and logging hundreds of incidents over the past year. Three companies are racing to rebuild code hosting for the agentic era with fundamentally different approaches: Cursor unveiled Origin, a Git-compatible platform designed for agent workloads with MCP support and built-in CI failure resolution; GitLab announced Project Switch (Next Generation SCM), which keeps Git protocol but redesigns the backend to allow server-side repository queries — delivering up to 50x faster agent task execution with 3x fewer tokens; and Zed introduced DeltaDB, the most radical departure, replacing Git's commit model entirely with a continuous stream of fine-grained deltas tied to agent conversations. Former GitHub director Brian Douglas argues that tokens — not commits or lines of code — are the right metric for the agentic era, and that owning the underlying model (as Cursor does with Composer) is what makes owning the broader developer stack viable.

   76

   10
4. 4
   Article

   The Register·11w

   GitHub nukes 70+ Microsoft repos, breaks CI/CD pipelines, following suspected worm infections

   GitHub removed over 70 Microsoft repositories after suspected worm infections, disrupting CI/CD pipelines. The malware, dubbed Miasma, is described as shape-shifting but consistently targets cloud secrets and credentials. The takedown highlights risks of worm-based supply chain attacks spreading through developer infrastructure on major code hosting platforms.

   73

   13
5. 5
   Article

   Lobsters·11w

   our workplace LLM mass delusion

   A frustrated employee shares their firsthand experience of what they describe as irrational AI hype in their workplace. Despite severe budget constraints — cancelled bonuses, dropped licenses, understaffed departments — the organization spent heavily on AI consultants, LLM workshops, ChatGPT and Copilot licenses. Recurring company-wide demos produced zero successful use cases across all teams. Presented use cases included asking ChatGPT how it 'feels,' using it to read a one-page lunch menu, and uploading suspicious emails to it. The author argues this reflects a mass delusion amplified by the Dunning-Kruger effect, where leadership chases AI hype while real operational needs go unmet, and that the speed of AI adoption exposes how organizational inertia is a choice, not an inevitability.

   71

   7
6. 6
   Article

   The Next Web·10w

   LinkedIn will now show recruiters which apps you actually use, not just which skills you claim

   LinkedIn launched Connected Apps, a feature that links real software usage to your profile by auto-generating activity-based descriptions from tools like GitHub, Replit, Descript, and Lovable. Unlike self-declared skills, these summaries are generated from actual usage data and cannot be manually edited by users. The goal is to give recruiters verifiable proof of tool proficiency rather than unverified claims. Users are notified when summaries are added or updated, and the feature is opt-in. If recruiters begin filtering by verified tool usage, professionals who don't connect their apps may find themselves at a disadvantage.

   60

   20
7. 7
   Article

   Windows Command Line·10w

   Intelligent Terminal 0.1.1 is here: bash support, new slash commands, and more customization

   Microsoft's Intelligent Terminal 0.1.1 is now available, bringing several notable improvements. Auto error detection and fix suggestions, previously limited to PowerShell, now work in Bash and WSL Bash. A new /fix slash command lets users manually request help for failed or misbehaving commands without waiting for auto-detection. A /model picker allows switching AI models mid-session. The agent pane is now a customizable Profile with font, color, and opacity options. Windows 10 (build 19041+) is now supported. The release also fixes session management issues, blank agent views after first login, stale tab state on pane close, and agent executable path resolution.

   59

   7
8. 8
   Article

   GitHub Blog·9w

   I automated my job (and it made me a better leader)

   A senior director at GitHub shares how she built 40 automations using the GitHub Copilot desktop app to manage the invisible overhead of senior leadership. Automations connect to her calendar, email, Slack, and GitHub repos via MCP integrations to handle meeting prep, daily triage, launch tracking, stale PR detection, and career journaling. She frames automations as an accessibility tool for AuDHD, narrowing the gap between good and bad executive-function days. The post includes a real example prompt (Stale Work Finder) and practical advice on starting with one high-friction task and iterating from there.

   42

   7
9. 9
   Article

   GitHub Changelog·10w

   Upcoming deprecation of Opus 4.6 (fast)

   GitHub is deprecating Opus 4.6 (fast) across all GitHub Copilot experiences — including Copilot Chat, inline edits, ask and agent modes, and code completions — on June 29th, 2026. Users and administrators should migrate to Opus 4.8 (fast) before that date. Copilot Enterprise admins may need to enable the replacement model via model policies in Copilot settings before it appears in the model selector.

   36

   3
10. 10
    Article

    GitHub Changelog·10w

    Repository switcher generally available in global navigation

    GitHub has made the repository switcher in global navigation generally available. Users can now click the chevron next to a repository name in the navigation breadcrumb to open a dialog, search for, and jump to another repository within their organization or owner namespace without leaving the current page. This feature is particularly useful for developers who work across many repositories.

    29

    3
11. 11
    Video

    Chris Titus Tech·12w

    Why I Left VS Code and Neovim for Zed Editor

    A developer shares their personal workflow after switching from VS Code and Neovim to Zed editor. Key highlights include Zed's fast launch speed, strong Vim keybinding support, JSON-based settings and keymaps (synced across Windows/Mac/Linux via Git), a custom task system for automating workflows like screenshot-to-markdown insertion, and integrated AI agent support with GitHub Copilot. The author describes Zed as a hybrid of VS Code and Neovim, praising its clean layout, parallel agent threads for project switching, and flexible theme system.

    29

    14
12. 12
    Article

    ASP.NET Blog·11w

    Join us for .NET Day on Agentic Modernization Livestream

    Microsoft Reactor is hosting a .NET Day on Agentic Modernization livestream on June 16, 2026 (4–8 PM UTC). The event targets engineering teams modernizing legacy .NET applications and covers GitHub Copilot-assisted workflows, building agentic experiences with Microsoft Agent Framework and Microsoft Foundry, and automating migration/deployment with GitHub Copilot for Azure (Azure MCP). Sessions include modernizing WinForms and line-of-business apps using Aspire without full rewrites, and adding AI-powered capabilities post-migration.

    27

    2
13. 13
    Article

    Tech Lead Digest·11w

    Reorgs happen

    A 13-year GitHub veteran who survived 25 reorgs shares a practical framework for navigating organizational restructuring. Key insights: reorgs are predictable baseline conditions in tech (roughly every 6 months), not crises. The author advises investing in portable relationships, documented work, and async communication as reorg insurance. Tactical advice includes letting dust settle before reacting, re-onboarding deliberately with a new manager, and using reorgs as opportunities to reset narratives, renegotiate scope, and shed legacy work. The core thesis is that reputation, relationships, and documented work outlast any org chart.

    19

    1
14. 14
    Article

    BleepingComputer·11w

    Hands on with Intelligent Terminal, an AI-powered Windows Terminal

    Microsoft has released an open-source fork of Windows Terminal called 'Intelligent Terminal' that integrates AI agents directly into the terminal experience. It supports multiple AI models including GitHub Copilot, Claude, Codex, and Gemini. Key features include automatic error detection and suggestion (which sends failed commands to the selected AI for fixes), and session management that lets users resume previous agent sessions — something standard Windows Terminal doesn't support. The tool is available on the Microsoft Store and GitHub but is not bundled with Windows by default.

    19

    2
15. 15
    Article

    Where's Your Ed At·12w

    AI Doesn't Have ROI

    A sharp critique arguing that AI has no measurable return on investment and that the entire AI industry has been built on deliberately obscured costs. Key evidence includes Uber's COO admitting difficulty justifying AI spend, enterprises accidentally burning millions on Anthropic tokens, and GitHub Copilot's shift to token-based billing causing user outrage. The author contends that AI companies intentionally subsidized subscriptions to hide true costs, that AI job displacement is a myth unsupported by economic data, and that the AI bubble is fundamentally worse than the Dot Com Bubble because AI data center infrastructure — unlike fiber optic cable — will leave behind no reusable, cost-reducing legacy. The piece also dissects a SemiAnalysis report on 'dark output' and a Bain & Co. survey showing most enterprises see minimal AI cost savings, concluding that if AI truly worked as advertised, its economic impact would be impossible to miss.

    18

    9
16. 16
    Article

    Career Digest·12w

    GitHub Copilot's token billing shock, Wix cuts 20% of staff

    GitHub Copilot switched to token-based billing on June 1, with some developers reporting costs jumping from $29/month to $750 or more. Wix laid off ~1,000 employees (20% of staff), citing AI restructuring and currency pressures. Amazon shut down its internal Kirorank AI usage leaderboard after employees gamed it with unnecessary agent tasks — a Goodhart's Law problem also seen at Meta. Atlassian published data showing its Rovo Dev tool drove 19% more merged PRs per month across 3,400 repos and saved developers 2–3 hours per week. Salesforce confirmed zero engineering hires in FY2026, Morgan Stanley doubled its AI job loss forecast for European banks to 20% by 2030, and entry-level developer hiring is down 67% since 2022.

    16

    4
17. 17
    Article

    Collections·12w

    Microsoft releases Intelligent Terminal 0.1, an AI-integrated fork of Windows Terminal

    Microsoft Build 2026 delivered a broad set of developer tooling announcements. Key highlights include: the Windows Developer Config file (winget-based) going GA for one-command dev environment setup; 165+ Linux CLI utilities natively on Windows; WSL Containers (`wslc`) for running Linux containers without Docker Desktop; a git-aware File Explorer; Sample-based PGO with up to 33% performance gains; and a new WinApp CLI for packaging Windows apps. The most notable shell announcement is Intelligent Terminal 0.1 — an open-source, MIT-licensed Windows Terminal fork with native AI agent support that auto-detects command failures and suggests fixes via any ACP-compatible agent. Visual Studio updates embed AI into the debugger, profiler, and test runner, with bring-your-own-key (BYOK) support for enterprise compliance. GitHub Copilot gains a desktop app, an SDK for embedding agent loops in any app, and MAI-1 Code Fast — Microsoft's first from-scratch coding model. Hardware highlights include the Surface RTX Spark DevBox with 1 petaflop AI compute and 128GB unified memory capable of running a 120B parameter model locally.

    15

    9
18. 18
    Article

    .NET Blog·10w

    AI-Powered MSBuild Investigation with the Microsoft Binlog MCP Server

    Microsoft has released the Binlog MCP Server, a Model Context Protocol server that gives AI assistants like GitHub Copilot direct access to MSBuild binary log files (.binlog). It exposes 15 specialized tools covering build failure investigation, property tracing, performance analysis, and build comparison. Setup is available for Visual Studio (17.14+), VS Code, and CLI tools like Claude Code. The server is built on the MSBuild StructuredLogger library and supports natural language queries such as 'Why did my build fail?' or 'What's making my build slow?' Telemetry is on by default but can be opted out via DOTNET_CLI_TELEMETRY_OPTOUT.

    13
19. 19
    Article

    Thomas Thornton·12w

    Keep GitHub Copilot Agent Skills Small and Focused

    GitHub Copilot Agent Skills work best when kept small and focused on a single behaviour. Overloading a skill with too many instructions creates unpredictable output and competing priorities. The post recommends designing skills around one clear job, using specific descriptions so the agent loads the right skill, keeping SKILL.md concise while offloading deeper material to references/ directories, and using concrete examples over long rule lists. Scripts should only be added when repeatable validation is needed. The goal is consistent, predictable output that teams can rely on — not a mega-prompt trying to cover every scenario.

    12

    2