Community

A Study of Microsoft's Early 2026 Rollout of Claude Code and GitHub Copilot CLI

Kimi K2.7 Code is generally available in GitHub Copilot

GitHub Copilot App

Angry devs vow to flee GitHub Copilot as metered billing takes hold

AI costs how much? GitHub Copilot users react to new usage-based pricing system

"What a joke": GitHub Copilot's token-based billing spurs backlash among devs

GitHub Copilot silently inserts itself as a co-author

Tell HN: VS Code v1.117.0 automatically adds GitHub Copilot as your co author

I don’t even use copilot. Yet, if you are like me and sometimes use the Source Control tab to stage files or manually write commit messages, it automatically appends Copilot as the co author for that commit.<p>How does anyone at Microsoft&#x2F;Github thinks this is ok? I don’t even use your AI, this is code I have written myself. This just looks desperate.

GitHub Copilot code review will start consuming GitHub Actions minutes

GitHub Copilot is moving to usage-based billing and retiring annual plans

Hi there,<p>You&#x27;re receiving this because you have an annual Copilot Pro or Pro+ plan.<p>GitHub Copilot isn&#x27;t the same product it was a year ago. It has evolved from an in-editor assistant into an agentic platform capable of running long, multi-step coding sessions, using the latest models, and iterating across entire repositories. Agentic usage is becoming the default, and it brings significantly higher compute and inference demands.<p>To keep up with the way you use Copilot, we&#x27;re transitioning to usage-based billing, and we want to give you enough time to prepare. If you&#x27;re also an admin on a Copilot Business or Enterprise plan, you&#x27;ll receive a separate email covering what&#x27;s changing for your organization.<p>Last week, we made temporary updates to Copilot individual plans to improve reliability and performance ahead of the broader move to usage-based billing. We will loosen usage limits once usage-based billing is in effect.<p>What&#x27;s changing on June 1 for Copilot Billing<p>Premium request units, or PRUs, will be replaced by GitHub AI Credits which are a monthly allotment of credits consumed based on token consumption (input, output, and cached tokens) according to the listed API rates per model. This change aligns Copilot pricing with actual usage and is an important step toward a sustainable, reliable Copilot business and experience for all users. Copilot code review’s agentic architecture runs on GitHub Actions. Copilot code review will now consume GitHub Actions minutes, in addition to GitHub AI Credits. These minutes are billed at the same per-minute rates as other GitHub Actions workflows. Fallback experiences will no longer be available. Under the new model, usage will instead be governed by available credits. What&#x27;s changing on June 1 for Copilot Annual Plans<p>GitHub is retiring Copilot annual plans. As a current annual subscriber your Copilot Pro or Pro+ plan will continue as-is with premium request-based pricing. When your annual plan ends, your account will automatically transition to Copilot Free, which you can continue using at no cost. The experience on annual plans will change significantly: model multipliers will increase, and standard-tier models (currently 0x) will no longer be available, reflecting increased compute costs and the transition to usage-based billing, and no new models or features will be added to annual plans going forward. To keep getting new models and features, we recommend switching to a Copilot Pro or Pro+ monthly plan with the new usage-based billing system starting June 1. If you switch before your annual subscription expires, you&#x27;ll receive prorated credits for the remaining value of your annual plan.<p>If you switch to a monthly plan, Copilot Pro is $10&#x2F;month and Pro+ is $39&#x2F;month, and each includes $10 and $39 in monthly AI Credits, respectively.<p>You can find more details on usage-based billing in our blog post and FAQ. We appreciate your patience as we build a more sustainable Copilot for everyone.<p>The GitHub Team

GitHub Copilot is moving to usage-based billing

GPT-5.5 is generally available for GitHub Copilot

Announcement: Changes to GitHub Copilot Individual Plans

Changes to GitHub Copilot individual plans

MS to Shift GitHub Copilot Users to Token-Based Billing, Reduce Rate Limits

GitHub Copilot rate limits Pro Users

MiniMax M2.5 is now available in GitHub Copilot for free and student use

GitHub Copilot now lets Free users and students use MiniMax M2.5, but no official clients—Copilot CLI, VS Code, github.com/copilot UI, or OpenCode—display it.  However, it's in the model listing API 3 times and and usable via GitHub OAuth. The model listing API provides a lot of detail about it.  It indicates that M2.5 is being served by Fireworks, and that there's a "(Fast)" version served by Cerebras.   Both are only available to "free" and "edu" subscriptions, to compensate, one would think, for the removal of premium models from the student plan (https://github.com/orgs/community/discussions/189268).) Both are marked as preview, "model\_picker\_enabled=false" and "model\_picker\_category=powerful".  The ID of "MiniMax M2.5 (Fast)" is minimax-m2p5-cb instead of minimax-m2p5-fw. { "billing": { "is_premium": true, "multiplier": 1, "restricted_to": ["free", "edu"] }, "capabilities": { "family": "minimax-m2p5-fw", "limits": {"max_context_window_tokens": 196608, "max_output_tokens": 32000, "max_prompt_tokens": 164000}, "object": "model_capabilities", "supports": { "parallel_tool_calls": true, "reasoning_effort": ["low", "medium", "high"], "streaming": true, "structured_outputs": true, "tool_calls": true }, "tokenizer": "o200k_base", "type": "chat" }, "id": "accounts/msft/routers/mp3yn0h7", "is_chat_default": false, "is_chat_fallback": false, "model_picker_category": "powerful", "model_picker_enabled": false, "name": "MiniMax M2.5 (Copilot)", "object": "model", "policy": {"state": "enabled", "terms": ""}, "preview": true, "supported_endpoints": ["/chat/completions"], "vendor": "Fireworks", "version": "accounts/msft/routers/mp3yn0h7" } The reason the official clients don't show the MiniMax models is because the model_picker_enabled field on them is set to false.  In Piebald we weren't aware of that field, and therefore by accident we don't respect it, thus revealing these options.  They must be pretty new because we have plenty of Copilot users and none of have used MiniMax with Copilot before.

Is this a joke? weekly rate limit of 264 hrs (11 days)?

https://preview.redd.it/zflkaaam44vg1.png?width=462&format=png&auto=webp&s=658a1bae15dcf7428b6c993b2349fa7db4a858d1 Whats going on copilot team? \[Update 1\]: Tried after an hour - "Wait for 264 hrs" \[Update 2\]: Another try after some time - "Wait for 278 hrs" \[Update 3\]: Another try after some time - "Wait for 286 hrs" \[Update 4\]: After almost 18hrs - "Wait for 293 hrs" Copilot team wants me to take some break and go on a long vacation.. (hope I'd be sponsored) \[Update 5 - 5:45pm IST\]: "Wait for 301 hrs" And whats interesting is that for every such message response it costs credits, so i am trying with 0.33x models to say "hi" LOL \[Update 6 - 11:00pm IST\]: "You have reached your weekly rate limit. Please upgrade your plan or wait for your limit to reset on April 28" \-> wow weekly limit resets after 2 weeks. should rephrase the msg. \----------------- Usage is spread across multiple models, even including gemini 3.1, Sonnet but mainly gpt 5.4 and well within limits (15% so far should have been 50% so already way less usage), context managed as much possible, short tasks. So what they are saying has nothing to do with its, I believe its all about the shortage of their resources as the number of users have increased. And its a BUG, you cant have 300 hrs wait time for weekly limit .. or wait is it monthly limit.. but surely 15% usage wont be for monthly limit as well.

Pro+ Github blocked me for 2 days because I used their service

I own Github Pro+ Since February and it all was working fine. Not not only they removed Opus Fast, and "reduced limits on users" to only increase limit and now I cannot do work for 2 days? I specially reviewed their ToS that they so often quote for me to check, THERE IS NOTHING IN THERE that says rate limit, Github maybe you will enlighten me how can I know If I am going too fast so that I can Plan for this? I got 71% of requests left for the month, I won't be able to use them probably...

A long-awaited Feature is here!

GitHub Copilot CLI now supports remote sessions! https://github.blog/changelog/2026-04-13-remote-control-cli-sessions-on-web-and-mobile-in-public-preview/

​I built a transparent Mario game that runs OVER my IDE so I can play while waiting for Copilot to write code.

When you're vibe coding and letting AI do the heavy lifting, the downtime gets boring fast, and continuous work is a real pain in the ass. To get a little relaxation, I built a stress-relief mini-game that lives directly on your Windows desktop as a transparent overlay, letting you stomp Goombas right on top of your open IDE. You can download the setup file from my GitHub (https://github.com/bxf1001g/desktop_mario/releases) and let it run silently in the background; just hit Ctrl+Alt+Shift+M to instantly bring the game up, press Esc to quickly hide it when you need to look busy, and hit Ctrl+Alt+Shift+M again whenever you need another break. This isn't the full version yet, just something fun I threw together as a side thing, so contributors are always welcome!

CoPilot Pro + VSCode extension is kinda a better deal than I expected: so far I vastly prefer GPT 5.4 Extra High to Claude Opus 4.6 and I'm only at 6% usage after at least like five hours of heavy work with it

CoPilot Pro + VSCode extension is kinda a better deal than I expected: so far I vastly prefer GPT 5.4 Extra High to Claude Opus 4.6 and I'm only at 6% usage after at least like five hours of heavy work with it

Goodbye trial accounts

https://github.blog/changelog/2026-04-10-pausing-new-github-copilot-pro-trials/ Free stuff is over. Personally I think this a great idea. If you cant afford ten bucks for some machine to do your thinking, you're doing it wrong. Rate limiting might go away if all the leeches are removed.

AI is making mediocre engineers harder to spot

Not a hot take. Just something I’ve been noticing lately. Everyone on my team uses AI now. Code, infra, debugging, even architecture ideas. Productivity is definitely up. But… there’s a weird side effect. \--- Case 1 — trying everything, fixing nothing A guy was debugging a slow endpoint. Asked AI → got a bunch of suggestions: \- add caching \- batch requests \- async processing He tried all of them. Still slow. Turned out the query was missing an index. That’s it. The problem wasn’t that AI was wrong. It just wasn’t the right question. And if you don’t even know “missing index” is a thing to check, you’re basically guessing — just faster. \--- Case 2 — sounds right, breaks in real life Another one: someone built a rate limiter based on AI suggestions. AI said: “store counters in memory for performance”. Which… yeah, makes sense. Until you deploy multiple instances and everything falls apart. Now your rate limit is basically random. Again, AI didn’t lie. It just didn’t know (or wasn’t told) the real constraints. \--- That’s the pattern I keep seeing AI doesn’t make engineers worse. It just makes it easier to: \- look like you know what you’re doing \- ship something that “seems fine” \- and completely miss the actual problem \--- The scary part? These people look productive. \- PRs are clean \- features ship fast \- infra “works” But ask one level deeper: \- why this approach? \- what’s the trade-off? \- what happens under load? …and things get very quiet. \--- To be clear — I use AI every day I’m not anti-AI at all. It’s insanely good at: \- boilerplate \- exploring options \- explaining stuff quickly \- getting you unstuck But it’s not the one: \- making the final call \- understanding your system \- taking responsibility when things break That’s still on you. \--- Feels like the bar is shifting Before: \- you had to know stuff to build things Now: \- you can build things without fully understanding them And that gap only shows up when: \- something breaks \- or someone asks the “why” questions \--- If there’s one thing I’m trying to avoid right now: Becoming someone who can ship fast… but can’t think deeply. \--- Anyway, curious if others are seeing the same thing Is AI actually making us better engineers? Or just faster ones?