Find the best tweets about GitHub Copilot, from coding workflows and agent features to reviews, limitations, and developer productivity. Updated every two weeks.
Specific Copilot workflows, releases, comparisons, and engineering lessons from active developers.
Original Xholic analysis
Across 50 posts from 41 creators, developer conversation on GitHub Copilot is predominantly favorable: 78% of posts are SUPPORTIVE in stance and 72% carry positive sentiment. Reach is concentrated in platform news β the Ollama integration in VS Code (score 766.63, 70.27x the 10.91 all-time median) and the Copilot SDK announcement (205.17) β and themes cluster on model/platform releases (16%), Copilot CLI agentic workflows (14%), and custom agents/context engineering (12%). Separate from the stance split, the most prominent critical threads are the Copilot agent 'product tips' injected into pull requests, the default-on training-data change, and cost jumps reported after the token-billing swi
72% of posts
All-time engagement
44% of posts
Published in 90 days
Conversation map
Comparisons and market-share shifts against Cursor, Claude Code, Codex, and Claude Enterprise: Ramp spend data, JetBrains adoption survey, DIS dropping Copilot, 'Copilot will beat Claude Code' enterprise-inertia takes, tool-switching and 't
16%
Model drops in Copilot (Grok 4.5, Claude Opus 4.7, Sonnet 4.6), Ollama/local model integration in VS Code, plan availability and admin enablement, Copilot App GA with canvas.
16%
GA launch of Copilot CLI, power-user shortcuts (/compact, autopilot toggle, preToolUse hooks, one-shot prompts), the /pr command for CI fixes and review feedback, self-written TypeScript extensions with hot reload, referencing Issues/PRs, Y
14%
Default-on training on user interaction data starting April 24 with opt-out guidance, GDPR/LGPD concerns, the Copilot agent 'product tips'/ads injected into 1.5M pull requests, GitHub's 'programming logic issue' response and trust fallout.
14%
Embedding the Copilot CLI engine into apps via the SDK: multi-language support, planning loops, tool invocation, MCP integration, BYOK, replacing hand-rolled orchestration plumbing.
12%
GitHub's study of 2,500+ instruction files (context problem vs capability problem), copilot-instructions.md, path-scoped instructions, .agent.md personas, org-level custom instructions, agent skills, multi-agent parallel fleets, and Skill-
12%
Figma MCP roundtrip between code and canvas, design-to-code generation, Work IQ team context in CLI, seamless session handoff between CLI and VS Code.
8%
Token-based billing switch and bill-shock stories ($29β$750+), per-session credit caps and spend-limit practices, plan mode and trust-tiered approvals, enterprise economics debates (Copilot deemed too expensive, flat-rate AI declining).
8%
Tone and stance
Performance benchmark
Posts with media make up 62% of this collection. Their median all-time score is 12.8, compared with 7.62 for text-only posts.
Format mix
Consensus and debate
Shared view
Multiple independent posts describe the Copilot CLI as a full agentic loop in the terminal: the GA announcement, self-written TypeScript extensions with hot reload in experimental mode, the /pr command for CI fixes and review feedback, and power-user shortcuts (/compact, autopilot toggle, preToolUse hooks, one-shot prompts).
Shared view
Two posts describing GitHub's study of 2,500+ instruction files converge on the same conclusion: agents have a context problem, not a capability problem. Effective setups pair personas, exact commands, boundaries, and output examples; Copilot's layered system (copilot-instructions.md, path-scoped instructions in .github/instructions/, .agent.md custom agent personas, org-level inheritance) is cited as the concrete
Shared view
Posts consistently describe the SDK as exposing the production-tested agent runtime behind Copilot CLI for programmatic use β planning loops, tool invocation, MCP integration, BYOK β replacing hand-rolled orchestration plumbing in applications.
Open debate
Opinions diverge sharply. One post argues Copilot will beat Claude Code on enterprise inertia rather than quality ('not because it's better'); another reports companies choosing Copilot over Cursor and Claude; a third finds it noticeably worse than Codex and Claude Code after weekend comparison. Market-share data points in both directions too: Ramp spend data reportedly shows Cursor overtaking Copilot in business
Open debate
One post reports a former colleague's team deeming $19/dev/month Copilot too expensive and hunting for a cheaper agent; another describes Copilot's Opus quota as much larger than Claude Code's and reports a one-prompt integration test succeeding without intervention. The token-billing switch (June 1) sharpens this split, with reported bill jumps from $29/month to $750 β some as high as $3,000.
Open debate
Posts differ on scale and framing: one reports promotional messages on more than 11,000 pull requests; another reports over 1.5 million pull requests containing some form of Copilot advertisement at 1,000+ injections per day. A GitHub response post attributes it to a 'programming logic issue' introduced March 24 and says agent tips have been removed from all pull requests; a Microsoft statement reported by Windows
What performs
ANNOUNCEMENT-format posts have a median all-time score of 15.92 versus 4.52 for OPINION posts, and posts with media have a median score of 12.76 versus 7.62 for text-only posts. The single CASE_STUDY post (score 102.72) about the Copilot runtime rewrite from TypeScript to Rust β one engineer, ~3 weeks, $120K in credits, 800K lines, up to 21x faster in-process β outperforms nearly everything except the largest
Five posts sit far above the 10.91 median: the Ollama/VS Code integration (766.63), the Skill Recorder agent-tooling post (511.4), the Copilot SDK (205.17), the paused Pro signups/compute-demand post (157.91), and the Figma MCP roundtrip (105.81). Notably, two of the five are critical or cautionary posts, showing reach is not confined to favorable content.
Median post age is 215.2 days, and only 4 of 50 posts (8%) were published within the last 90 days, with posts spanning 2025-06-23 to 2026-10-24. Evergreen workflow and theme content dominates; genuinely recent posts are a small minority.
The TUTORIAL format (20% of posts) has a median score of 6.62, with the strongest performers being the GitHub instruction-file study breakdown, the freeCodeCamp piece on replacing Copilot with Claude Code for Python/React projects, and the Copilot CLI cheatsheet post.
Statistical standouts
Creator landscape
The five most represented creators account for 20% of the selected posts.
1. Dan Wahlin
@DanWahlin
2 posts
2. GitHub Changelog
@GHchangelog
2 posts
3. John Crickett
@johncrickett
2 posts
4. Layton Gott
@Layton_Gott
2 posts
5. Martin Woodward
@martinwoodward
2 posts
6. Priyanka Vergadia
@pvergadia
2 posts
Satya Nadella's two posts have a median all-time score of 133.09, including the Copilot SDK announcement (205.17) and the Work IQ team-context post for Copilot CLI (61.01) β the highest per-post reach of any listed creator.
John Crickett's two posts have a median all-time score of 43.49, both on the competitive/enterprise side: the 'Copilot will beat Claude Code' enterprise-inertia take and the '$19/dev/month deemed too expensive' economics post.
Martin Woodward (median 13.61) spans both the PR product-tip response and the CLI-to-VS Code session-handoff experience; Dan Wahlin (14.05) covers the Copilot App GA/canvas and CLI getting-started content; GitHub Changelog (6.82) posts org custom instructions and custom agents/skills in Visual Studio. The creator pool is broad: 41 creators, with the top five voices accounting for 20% of placements.
Since the previous snapshot
Themes, sentiment, stance, and post format are classified per tweet. All counts, shares, medians, creator concentration, freshness, and performance comparisons are then calculated directly from the published snapshot.
Xholic's all-time score compares engagement while accounting for reach, post age, and creator consistency. It is used for relative comparisons within this collection.
This report analyzes the exact 50-post snapshot shown below. AI identifies editorial categories and drafts explanations; all statistics are calculated from the snapshot, and every narrative claim is checked against cited posts before publication.
Best GitHub Copilot tweets
Ranked 01β50
@ollama Β·
Visual Studio Code now integrates with Ollama via GitHub Copilot. If you have Ollama installed, any local or cloud model from Ollama can be selected for use within Visual Studio Code.
@RoundtableSpace Β·
Microsoft open-sourced an app that watches you do a task once, then teaches your AI agent to do it for you. 4.3K stars. How Skill Recorder works: 1. Hit record (ββ§R or Ctrl+Shift+R) and do the task normally. You can talk through it out loud too. 2. It captures your clicks, app switches, pages and terminal commands, all on your machine. 3. GitHub Copilot turns the recording into a list of steps you can edit. 4. One click turns it into a skill your agent can run, or an automation on a schedule. It generalizes from one example, so recording yourself submitting one form teaches the agent to submit all of them. Runs on Mac, Windows 11 and Ubuntu. Needs a GitHub account with Copilot access. Install: copy the one-line command from the latest release: https://t.co/NmDteJ1WgB
@satyanadella Β·
A new developer workflow and app paradigm is emerging, with an agentic execution loop at the core. With the GitHub Copilot SDK, you can embed the same production-tested runtime behind Copilot CLIβmulti-model, multi-step planning, tools, MCP integration, auth, streamingβdirectly into your apps. https://t.co/RamJvw2U9D
@trikcode Β·
So GitHub Copilot just paused new Pro signups because they ran out of compute. the biggest AI coding tool on the planet literally cannot handle the demand. the AI infrastructure crisis is here
@figma Β·
Github Copilot to Figma: Roundtrip between code and canvas through the Figma MCP server
Watch video
@mjovanovictech Β·
Microsoft moved the GitHub Copilot runtime from TypeScript to Rust with AI agents. One engineer, ~3 weeks of dev time, $120K in credits, 800K lines. Result: up to 21x faster in-process. Nobody read every line. Here's how they reviewed it: https://t.co/uhf4UUjJkM
@akshay_pachaar Β·
There's an interesting study by GitHub on coding agents! They analyzed 2,500+ custom instruction files across public repos to understand what separates effective agent setups from weak ones. Effective setups give agents a specific persona, exact commands to run, defined boundaries, and examples of good output. Weak ones are vague helpers with no clear job description. This points to the core friction with coding agents today, which is that they don't have a capability problem but rather a context problem. A raw agent can write code, but it doesn't know the team's naming conventions, the specific linting setup, or preferred framework patterns. Without that context, the first PR is often off-target and requires multiple rounds of correction. Getting this right requires structured context, and GitHub Copilot implements a smart, layered customization system that does exactly this. > At the repo level, a `.github/copilot-instructions .md` file defines project-wide rules like coding conventions, naming standards, security defaults, and prohibited patterns. The agent reads this before generating any code. > For granular control, instruction files in .github/instructions/ can target specific file paths using applyTo frontmatter. A TypeScript-specific instruction file only activates when the agent works on .ts files. > The most interesting addition is custom agents. These are `.agent .md` files in `.github/agents/` that define specialized personas with their own tool access and MCP server connections. For instance, a security auditor agent can be configured with only read access and run linters before flagging issues. A test writer agent can follow specific testing patterns defined by the team. Each agent has defined boundaries for what it can and cannot do. These custom agents can also be defined at the organization level in a .github-private repo and inherited across all repositories. Frontend conventions, backend patterns, and security policies apply everywhere without duplicating config files. But the customization doesn't stop at DIY setups. There's more π
Watch video@satyanadella Β·
So much of dev work happens in the context of a larger team, and now you can bring all of that work context into the GitHub Copilot CLI with Work IQ.
Watch video
@cb_doge Β·
BREAKING: Grok 4.5 is now available in GitHub Copilot Built for fast agentic coding and complex multi-step workflows, it offers a massive 500,000-token context window, image support and adjustable reasoning effort. Available across VS Code, Copilot CLI, JetBrains, Xcode & more. https://t.co/m1zLFnzPDl
@johncrickett Β·
Chatting with a former colleague today: - Their CEO has mandated 80%+ of code must be written by AI. - GitHub Copilot at $19/dev/month is deemed too expensive. - Teams are being told to find a cheaper agent. The economics of AI adoption are getting interesting.
@veggie_eric Β·
Grok 4.5 is live in GitHub Copilot! Available for Pro, Pro+, Max, Business, and Enterprise plans; ask your admin to enable the model in Copilot settings.
@Pirat_Nation Β·
GitHub Copilot added promotional messages to more than 11,000 pull requests on GitHub. A developer asked Copilot to fix a simple typo in a pull request. Copilot made the fix but also added this line to the description: "Quickly spin up Copilot coding agent tasks from anywhere on your macOS or Windows machine with Raycast." GitHub called it a programming logic issue.
@johncrickett Β·
GitHub Copilot will beat Claude Code. Not because it's better. Because Copilot is the new IBM. It checks the enterprise box. Microsoft is already a preferred supplier. And if it under-delivers, well everyone else bought it too. "You won't get fired for buying X" really means "you won't get blamed for buying X." There's a big difference between not getting blamed and making the right call.
@_Evan_Boyle Β·
Copilot CLI can now write and hot reload it's own typescript extensions! Here the agent wrote it's own snake extension that only lets me play while copilot is busy, and pauses the game when I need to provide input. Available today in /experimental mode v1.0.3
Watch video
@RaulJuncoV Β·
GitHub Copilot switched to token-based billing on June 1. Devs are reporting costs jumping from $29/month to $750. Some saw $3,000. Code completions are still unlimited. But everything else? You're paying per token. Your GitHub Copilot just became a taxi with the meter running.
@freeCodeCamp Β·
Autocomplete speed matters more than you may realize when you're coding all day. In this article, Balajee shares what happened when he replaced GitHub Copilot with Claude Code on his Python and React projects. He discusses where Claude Code felt smarter, where latency hurt productivity, and why using both tools worked better in the end. https://t.co/ys29jcPcn0
@HedgieMarkets Β·
π¦ GitHub Copilot has been injecting ads into pull request descriptions at scale. Over 1.5 million pull requests on GitHub now contain some form of Copilot advertisement, with more than 1,000 injections per day over the past ten days promoting tools like Raycast, Jira, Linear, VS Code, and JetBrains. A GitHub team member confirmed the practice on Hacker News and said they have now disabled it. It is not yet clear whether this was an intentional monetization test or something that got out of hand. My Take A pull request is a developer's technical contribution to a software project. It is not an ad unit. Microsoft tested injecting promotional content into 1.5 million of them without announcement and only stopped when it went public. That sequence is the part worth paying attention to. Microsoft is having its worst start to a year this century, froze hiring in cloud and sales last week, and is carrying billions in OpenAI losses that need to be offset somewhere. When the core product starts looking for ad revenue inside developer workflows, that is not a sign of confidence in the underlying business model. The more pointed question, raised by several developers already, is straightforward. If Copilot is already comfortable modifying the metadata of your pull requests without telling you, what else is it comfortable modifying without telling you. Hedgieπ€
@martinwoodward Β·
Hey folks, had some conversations about this today so thought it would be helpful to be clear. GitHub does not and does not plan to include advertisements on the platform. We recently identified a programming logic issue in the GitHub Copilot coding agent that caused a 'product tip', including a third-party suggestion, to appear incorrectly in a pull request comment. This issue was introduced on March 24 during a rollout that expanded Copilotβs ability to contribute to any pull request when requested to by a developer. As a result, a third-party link was mistakenly displayed in a way that could be interpreted as a promotion. Our goal was to share novel ways to use Copilot coding agent, and in this case, we highlighted our integration with Raycast as part of a broader set of product tips, but this was surfaced more frequently than intended alongside other feature suggestions. We have removed Copilot agent tips from all pull requests moving forward. We appreciate the community flagging this and apologize for the error.
@amandaksilver Β·
GitHub Copilot CLI is now GA π This is agentic development, right in the terminalβplanning, building, testing, and reviewing without context switching. A big step forward in how developers work. π https://t.co/6PbrLEu7tI
@mariorod1 Β·
One of Sonnet 4.6 greatest improvements, based on our early testing, is its ability to search, navigate and resolve issues across large codebases. Available in GitHub Copilot now π. https://t.co/SYPIXeQvVK
Watch video
@arakharazian Β·
Last month @cursor_ai overtook GitHub Copilot in business spend, @tryramp data shows. Both continue adding users + spend, more than enough to go around in this market. But goes to show that first movers != market dominance. Not included: Claude Code, small but growing
@DanWahlin Β·
The GitHub Copilot App is now GA and it has a canvas feature that's amazing! π Here's how I integrated Agent Arcade into it. πΉοΈ GitHub Copilot App: https://t.co/JQ6dNFcWX9 Canvas Documentation: https://t.co/4hs46lKLic Agent Arcade Repo: https://t.co/LG9S23vz6k Agent Arcade Canvas Extension on Awesome Copilot: https://t.co/0JqgDvIsuv Shoutout to @spboyer for the initial PR!
Watch video
@pvergadia Β·
NEW: 5 @github Copilot CLI shortcuts that save hours. Most devs don't know they exist. Copy these now: β -p "fix this bug": one-shot prompt, no interactive session, perfect for scripts β Shift+Tab: toggle autopilot mode. Agent runs without asking permission. Every. Single. Time. β /compact: context window filling up? Summarize history and keep going. No restart. β preToolUse hook: intercept any tool call before it fires. Modify args. Block writes. Live. β !COMMAND: run raw shell mid-session without breaking context My full GitHub Copilot CLI Cheatsheet π
@WindowsLatest Β·
Microsoft denies rolling out Copilot ads on GitHub after pull requests were found with ad-like notes, causing outrage. In a statement to Windows Latest, Microsoft confirmed that the Copilot-powered ad-like notes were not paid advertisements, but the result of a logic issue. Microsoft has also confirmed it wonβt add ads to GitHub. "GitHub does not and does not plan to include advertisements in GitHub," Microsoft told Windows Latest. "We identified a programming logic issue with a GitHub Copilot coding agent tip that surfaced in the wrong context within a pull request comment. We have removed agent tips from pull request comments moving forward."
@tiagonbotelho Β·
The new /pr command in the Copilot CLI handles the parts of SDLC that nobody enjoys. Fixing failing CI, addressing review feedback, resolving merge conflicts. Now you can just focus on shipping πΏοΈ
Watch video@tiagonbotelho Β·
The team's been cooking with this one. You can now filter and reference Issues and PRs straight from the Copilot CLIβ¨
Watch video@DanWahlin Β·
Skip context-switching and let AI build stuff right in your shell! β¨ In this video you'll learn how to get started with GitHub Copilot CLI - from quick 'Hello World' scaffolds to planning & building a weather app that gets live data from an API. π₯ https://t.co/eRCcRySxi2 Want even more? π
@GHchangelog Β·
Organization custom instructions for GitHub Copilot are now generally available. β’ Admins for Copilot Business & Enterprise can set default instructions across all org repos. https://t.co/QmRGPn6F4y
@itsafiz Β·
This is huge news! GitHub Copilot SDK made it easy to add agents to your application. The GitHub Copilot SDK is now available in Technical preview as a programmable SDK. Python, TypeScript, Go, and .NET It exposes the same engine behind Copilot CLI: a production-tested agent runtime you can invoke programmatically. No need to build your own orchestrationβyou define agent behavior, Copilot handles planning, tool invocation, file edits, and more.
@Shruti_0810 Β·
Anthropic just flipped the switch on coding AI. Claude Opus 4.7 inside GitHub Copilot isnβt an upgrade. Itβs a replacement mindset. AI no longer suggests codeβ¦ It finishes the job. β Multi-step tasks donβt collapse β Agents actually execute β Context doesnβt die mid-way This is the shift: Less prompting. Less fixing. More shipping. Most devs will use it like 2024 toolsβ¦ and stay stuck. A few will push it β and realize coding just changed forever.
Watch video@GHchangelog Β·
GitHub Copilot in Visual Studio now supports custom agents and reusable agent skills to enhance automation and code understanding. β’ Admins can enforce MCP server allowlists for better security and compliance. https://t.co/6M5ceg5690
@adisingh Β·
Most surprising thing I learned at AIE- Apparently GitHub Copilot⦠is pretty good. Good to the point where companies roll with it over Cursor and Claude. Yes, you read that right.
@pvergadia Β·
GitHub studied 2,500+ coding agent instruction files and found something that flips the narrative. Agents don't have a capability problem. They have a context problem. β Weak setups: vague helpers with no job description β Strong setups: specific persona, exact commands, defined scope, output examples β GitHub Copilot's fix: a layered system β repo-level rules, path-specific instructions, custom agent personas β A .github/copilot-instructions.md file runs before the agent touches a single line β .agent.md files create specialists: a security auditor with read-only access, a test writer locked to team patterns The first PR is always off-target when the agent doesn't know your naming conventions. Raw capability is now a commodity. The teams that ship clean code with agents aren't running smarter models β they're writing better context files.
@rohanpaul_ai Β·
New Microsoft Paper on GitHub Copilotβs production traces show why coding agents should not be served like chat requests. In 13.5M GitHub Copilot sessions, 87% of LLM calls came from the agent itself rather than a user. A user prompt can fan out into an autonomous chain of model calls, tool actions, retries, and growing context, making the turn or session a more useful scheduling unit than an isolated request. That structure is especially visible in the KV cache. It shows that KV cache is not really a request-level resource; its value depends on where the agent is in the workflow. Within a turn, average cache hit rate rises from about 45% on the first LLM call to 92β94% from the third call onward. At a same-model turn boundary it falls to 55%, while a model switch pushes it down to 8%. Median KV-cache idle time is 1.2 seconds within a turn versus 172 seconds across turns, while container idle time jumps from 5.8 to 243 seconds. Using turn- and session-level features, the paperβs lightweight predictor captures 86β90% of total idle time, giving the serving stack a signal for cache offloading or container reclamation. The implication is straightforward: coding-agent infrastructure should schedule workflow state across turns, because request-level policies discard some of the strongest signals in the workload. β arxiv. org/abs/2608.00101 Title: "Agentic Coding in the Wild: Characterizing GitHub Copilot Traces at Production Scale"
@vlad_mihalcea Β·
Thanks, @brunoborges, for the GitHub Copilot Pro subscription. It works like a charm. While I use Claude Code for my PhD research, I don't use it for coding as I'd quickly run out of tokens. However, Copilot can also use Opus, and the quota is much larger than for Claude Code. For example, I told Copilot to create an integration test for Camunda (this is a new dependency in this repo, and there is no example to use as a reference) with JTA and Spring, while reusing the DataSourceProvider mechanism employed by all tests in my High-Performance Java Persistence repository. One prompt, and job done. I didn't have to intervene at all, and the test works like a charm. Awesome!
@ghumare64 Β·
Another big drop by Microsoft in 2026. GitHub Copilot is truly open to the world. You can build your own Copilot using the Copilot SDK starting today. Thank you @github
Watch video
@Layton_Gott Β·
You can run coding agents with zero limitsβ¦ A lot of people do. Yet they get shocked by the cost because they used it recklessly. Limitless setups are insanely powerful but only when used right otherwise youβll be spending a fortune. So here's the cost saving breakdown: Start with a hard spend cap per task. GitHub now lets you set a per session credit limit on Copilot CLI and SDK jobs. The cap covers the model calls, the subagents, and the background stuff. So an agent physically cannot spend past the number you set. If your tool has this, turn it on. If it doesn't, build your own tool that watches the spending. Every autonomous task should have 6 limits: β’ Max spend β’ Max time β’ Max retries β’ Which tools it's allowed to touch β’ When it stops β’ When it asks you before continuing Next, approve the plan, not every action. Anthropic built Plan Mode for exactly this. Instead of clicking approve on forty tiny steps, the agent shows you the whole plan first and you approve the strategy once. But not everything should get the same trust level. Set it up like this: β’ Reading files, run automatically β’ Reversible local changes, approve the plan β’ Anything that talks to the outside world, approve it directly β’ Production changes, approve it directly β’ Payments or deleting things, separate confirmation on its own The mistake is giving a delete-database action the same easy approval as reading a file. This trick Iβve found to be the most cost saving: Don't load every tool and skill into the agent at once. Every tool you attach eats context and adds another thing that can go wrong. Big skill collections are spreading everywhere right now, and most of them are dead weight. They sit in context, cost tokens on every call, and rarely get used. The fix is on demand tool discovery. The agent pulls the tool it needs for the task instead of carrying all of them all the time. GitHub already has a version of this. It ranks and loads only what fits the current job. Test every skill you add with one question. Does it improve the completion rate enough to justify the tokens and the risk. If it doesn't, cut it. Last one: Stop measuring agents like chatbots. Message count means nothing for a long running agent. Anthropic literally had to rebuild their own research pipeline because chat transcripts stopped capturing how people actually use these things. The numbers that matter now: β’ Completed tasks β’ How often you had to step in β’ Accepted outputs β’ Tool errors β’ Cost per finished result β’ Rollbacks A model that looks cheap per token can be your most expensive one once you count the retries and the corrections. Put it together and the picture is simple. You wouldn't give a new hire unlimited spend, every key in the building, and no one checking in. Don't give an agent that either.
@catmanyau Β·
Give GitHub Copilot your app context. Ask for model recommendations, then compare reasoning, cost, latency, and context length. The useful shift, in my view, is making those tradeoffs explicit while the app is still taking shape, instead of choosing by name or instinct. A Copilot shortlist is a starting point, not proof of what will work best for your workload. Describe what your app needs in Copilot, use its shortlist to pick candidates, then evaluate them against your workload before committing.
Watch video@Layton_Gott Β·
GitHub Copilot devs are posting bill estimates going from $29/month to $750/month since the switch to token billing... A 25x increase for the same work. This is exactly why I maxed out my Claude subscription this week instead of waiting. Flat-rate AI is dying. Lock in the all-you-can-eat plans while they still exist, because every company is watching Copilot test how much usage-based billing devs will tolerate.
@sabir_huss50540 Β·
π¨BREAKING: GitHub just took the agent engine running inside Copilot CLI and handed it to you as an SDK. It's called the GitHub Copilot SDK. Everyone building an AI agent right now is writing the same plumbing from scratch: the planning loop, the tool-calling, the file edits, the permission handling. Weeks of orchestration code before the agent does anything useful. This is that engine, already built, already production-tested. Here's what it actually does: β Exposes the same agent runtime behind Copilot CLI, callable from your own code β Ships in six languages: Python, TypeScript, Go, .NET, Java, and Rust β Copilot handles planning, tool invocation, and file edits β you just define behavior β Define your own custom agents, skills, and tools on top of it β Speaks MCP, so external tools plug straight in β BYOK: bring your own OpenAI, Anthropic, or Azure key and skip GitHub auth entirely You stop writing the orchestration layer. You start writing the agent. The people shipping real agents this month aren't smarter than you. They just stopped rebuilding the runtime. 9.5k stars. MIT License. (Link in the comments)
@Techjunkie_Aman Β·
This is big. GitHub Copilot will train on your data. By default. Starting April 24: β’ Prompts + code you type β’ Suggestions (accepted, edited, rejected) β’ File context + repo structure β’ How you interact with the tool This isnβt about stored repositories. Itβs about your real-time workflow while coding. That data is used to improve models and make suggestions more context-aware. Default is ON. You have to manually opt out. How to opt out (takes ~10 seconds): β’ GitHub β Profile β Copilot settings β’ Find: βAllow data for AI trainingβ β’ Set it to Disabled If you donβt see it, youβre likely on a Business/Enterprise plan. Important: β’ Starts April 24, 2026 β’ Only affects interaction data (not stored repo code) β’ You can change it anytime Most people wonβt check this. Will you?
@shawnchauhan1 Β·
GitHub Copilot tried to put promotional content inside your code reviews this week. Developers noticed immediately. The feature was pulled within days. This tells you something important about where AI tools are heading. The free-to-cheap pricing phase was always temporary. The question was never whether monetization would come - it was where they would hide it. Burying ads inside a developer's most trusted workflow is the answer Microsoft tested. The backlash tells you the trust cost of that answer.
Watch video
@StockSavvyShay Β·
$DIS reportedly plans to drop $MSFT GitHub Copilot for U.S. staff in August and shift to OpenAIβs Codex. The move follows low Copilot usage and stronger employee adoption of rival tools while Claude Enterprise and Cursor will remain available.
@martinwoodward Β·
Being able to seamlessly pick up GitHub Copilot CLI sessions in VS Code and bounce between the two just feels like magic. One of those 'well of course it works like that' experiences that are so incredibly hard to pull off in how invisibly it all just works.
Watch video
@tristanbob Β·
I really want to like GitHub Copilot because it is the approved tool at work. But after using Codex and Claude Code all weekend, GitHub Copilot makes me sad. It's good, but noticeably worse. Codex is open source right? Why can't they just use that?
@akaclandestine Β·
Starting April 24, 2026, GitHub Copilot will begin using, by default, interaction data from users on the Free, Pro, and Pro+ plans to train and improve its artificial intelligence models. This change does not apply to customers on Business and Enterprise plans. The data covered includes: user inputs, generated outputs, code snippets, and the associated interaction context. Primary Official Source: https://t.co/phyfWE1Pz5 Complementary Update in GitHub Changelog: https://t.co/RtYeASf3Nw Recommended Mitigation Actions (Immediate Action Advised): 1. Access your GitHub account settings. 2. Navigate to Copilot > Privacy (direct link: https://t.co/T7yHP5jSNu). 3. Disable the option:β¨βAllow GitHub to use my data for AI model trainingβ. #GitHubCopilot #InformationSecurity #DataPrivacy #LGPD #GDPR #ArtificialIntelligence #IntellectualProperty #InfoSec #CyberSecurity #SecureDevelopment
@ThePracticalDev Β·
GitHub Copilot tackled an Electron refactor, fixed a long-standing OpenBSD bug, wrote shell tab-completion, and pushed 100% test coverage without being asked. This dev was skeptical, then very, very impressed. { author: @ David Cantrell } https://t.co/ckLTJQQdo3
@ThePracticalDev Β·
Running a fleet of Copilot agents in parallel means each one handles its own task with its own context leading to less drift and fewer tokens. This dev walks through Local mode, Copilot CLI, custom agents, and the tradeoffs of multi-agent execution. { author: @ThisIsIvoLondon } https://t.co/uWdVc9a1b3
@Afruturist Β·
According to the JetBrains AI Pulse Survey of 11,000 developers worldwide (January 2026), GitHub Copilot leads AI coding tool adoption at 29%, followed by ChatGPT used for coding at 28%. Claude Code and Cursor are tied in second place at 18% each, while Gemini, Claude Chat, Google Antigravity, and OpenAI Codex trail behind at 8%, 7%, 6%, and 3% respectively. Important - As developers commonly use multiple tools simultaneously, figures exceed 100%.
@dailydotdev Β·
GitHub Copilot can now pull designs from Figma and generate code no more "can you export those icons as SVGs?" back-and-forth your IDE finally speaks designer
Best GitHub Copilot tweets
Xholic studies what works in your niche, drafts posts in your voice and schedules them for the hours your audience is online.
No card needed Β· Nothing to cancel
Browse all tweet collectionsLive demo: no sign-up or login needed Β· takes about 2 minutes
Keep exploring