Alex Sergeyev’s Post

Cheap AI is ending. Faster than people think. In the last days: ↳ GitHub paused GitHub Copilot Pro and Pro+ signups ↳ Opus removed from the $10 GitHub Copilot plan ↳ Tighter limits across individual GitHub Copilot tiers ↳ Leaks point to token-based billing for all GitHub Copilot users in June ↳ Anthropic briefly dropped Claude Code from its $20 Pro plan (for some lucky users) I've been saying this for months. The $20 all-you-can-eat era was never going to last. The economics don't work. If you want to experiment and build real intuition with these tools, do it now. The window is closing. Later will only be more expensive.

Called it. A few weeks ago I noticed GitHub paused new Copilot signups and removed models. I said then: token billing is coming. It arrived. Got the official email this afternoon. What's changing on June 1: • Premium requests become GitHub AI Credits based on actual token consumption • Copilot Pro: $10/month = $10 in credits • Copilot Pro+ stays $39/month = $39 in credits • Code completions stay free, only agentic/chat features consume credits I'd already done the math on my April usage: • ~1,888 premium requests consumed • Gross value at API rates: ~$103 • What I paid: $39/month Under the new model, that same usage would cost me ~$103 in credits, on a $39 plan. So I'm now figuring out my actual options: - Stay on Copilot Pro+ and buy additional credits when needed - Claude Sonnet 4.6 direct API with prompt caching: ~$25–35/month realistic - GPT-5.4 Standard with caching: ~$20–30/month GitHub is launching a billing preview calculator in early May so I can see exactly what June would cost before committing to anything. I'll use that before deciding. In the meantime, I still have a month of nearly free tokens. The era of unlimited agentic AI on a flat subscription is over though. That part is certain. What's your plan for June? #buildinpublic #ai #softwaredevelopment #golang

GitHub Copilot is changing how Pro and Pro+ plans are billed starting June 1. The US$10 and US$39 monthly fees stay the same, but they will now work as monthly AI credits based on actual usage. • New model: billing based on input, output, and cache tokens. • Code Review: in addition to AI credits, it will also consume GitHub Actions minutes. • Autocomplete: remains included in the plan, with no credit usage. • Next Edit Suggestions: also remains included, with no credit usage. • End of fallback: once the limit is reached, there will be no fallback experience. In practice, GitHub is ending unlimited access for advanced features and aligning pricing with real infrastructure consumption. #GitHub #Copilot #SoftwareDevelopment #AI

🚀 What if your GitHub repo could improve itself? Join us for GitHub Agentic Workflows: Automation That Actually Reads the Room and see how AI agents can autonomously triage issues, fix CI failures, update docs, and improve tests—with just a markdown file and one command. 🔍 In this live demo, Ari LiVigni, Senior Learning Advocate at GitHub, will walk through: - From minimal workflow file → AI agent in action - Safe, sandboxed pipelines that deliver ready‑to‑review PRs - How to keep your repo on autopilot while staying fully in control 📅 May 7, 2026 ⏰ 9:00–10:00 AM PT 🎥 Live and on-demand at any time after wrap 👉 Register now: https://msft.it/6040QxqQE 📍 Explore the series: https://msft.it/6041QxqQ1 #MicrosoftReactor #learnconnectbuild #GitHubCopilot #AgenticWorkflows #AgentOrchestration #GHCP #Fleet

🔴 Going LIVE! Kayla Cinnamon, Senior Developer Advocate at Microsoft, is about to take the stage for her tutorial "GitHub Copilot Everywhere: CLI, VS Code, and the Cloud" at the Future of Data and AI: Agentic AI Conference! Join live: https://lnkd.in/dpNDdg74 If you're a developer looking to get the most out of GitHub Copilot across your entire workflow — this is the one to catch live. Real features, real trade-offs, no fluff. In this tutorial, you'll learn to: - Use GitHub Copilot across the CLI, VS Code, and cloud-based development workflows - Build real features end-to-end using Copilot across multiple surfaces - Understand the trade-offs between tools and when to switch - Apply practical patterns for integrating Copilot into your entire development loop #AgenticAI #FutureOfAI #AIConference #GitHubCopilot #DeveloperTools #Microsoft

🤖 One of GitHub's most interesting updates this week is happening in the CLI: GitHub Copilot CLI now supports BYOK and local models⚡ What that means: ▶️ teams can connect their own model provider ▶️ they can also run fully local models ▶️ GitHub-hosted routing is no longer the only option ▶️ authentication is optional when using your own provider Why this matters: ✅ more flexibility in model choice ✅ more control over LLM spend ✅ better fit for restricted or air-gapped environments ✅ easier alignment with enterprise infrastructure choices This is a meaningful step for teams that want the Copilot CLI experience, but with more control over how and where inference happens. #GitHub #GitHubCopilot #CLI #AI #DeveloperTools

GitHub 𝗽𝗮𝘂𝘀𝗲𝗱 𝗻𝗲𝘄 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝘀𝗶𝗴𝗻-𝘂𝗽𝘀 𝘁𝗵𝗶𝘀 𝘄𝗲𝗲𝗸 𝗮𝗻𝗱 𝘁𝗶𝗴𝗵𝘁𝗲𝗻𝗲𝗱 𝘂𝘀𝗮𝗴𝗲 𝗹𝗶𝗺𝗶𝘁𝘀 𝗮𝗰𝗿𝗼𝘀𝘀 𝗶𝘁𝘀 𝗶𝗻𝗱𝗶𝘃𝗶𝗱𝘂𝗮𝗹 𝗽𝗹𝗮𝗻𝘀. The reason they gave was direct: a handful of agentic sessions now routinely cost more than the plan price itself. That is not a GitHub problem. That is a structural one. Flat rate pricing was designed for a world where AI assisted. It was not designed for a world where AI acts. Extended, parallelized, long running agent sessions consume compute at a scale the original pricing models never anticipated. Limits will tighten. Models will disappear from lower plans. Access will be gated behind higher tiers. We are at the point where AI capability and AI cost have to be planned together. They cannot be treated as separate conversations anymore. Link: https://lnkd.in/ea-P7bxt

The subsidized AI compute wars continue to escalate at an more accelerated rate. GitHub Copilot removed Claude's Opus 4.5 & 4.6 from usage earlier and are restricting sign ups on their paid plans, red more here - https://lnkd.in/eKtXwSdM. Only Opus 4.7 remains at their increased request rate. Thankfully Sonnet 4.6 is still available for the less complex tasks Unfortunately they missed a few spots - Opus 4.5 & 4.6 are still listed as available on their docs as of 4/20 here https://lnkd.in/eZQtFj4v There are rumors that GHCP will shift to token-based billing, which has widely become the industry standard. This was expected, as Github Copilot's real advantage was allowing users to switch between a variety of models and prompt them for a fraction of the actual cost. I'm guessing fewer and fewer folks were prompting other models aside from Codex and a few others. All of these companies are heavily subsidizing this compute in an attempt to determine who the winner will be. To the winner goes the spoils and the majority of future tokens Sadly, this likely marks the end of any advantage GHCP's requests provided as it will most certainly be cheaper to purchase your tokens directly from Anthropic, OpenAI's, Cursor, and any other of the frontier labs Intrigued to see if Microsoft / Github have a model they've been working on behind the scenes and how this shift in strategy will impact their product and the greater landscape moving forward

🚨 GitHub Copilot is switching to usage-based billing on June 1, 2026. Instead of a fixed number of "premium requests," you'll get a monthly allotment of GitHub AI Credits. Why the change? Agentic workflows exploded Copilot's compute costs reportedly nearly doubling week-over-week since January. The flat-rate model simply isn't sustainable anymore 😉 . Interesting to compare with Claude Code, which uses a similar token-based philosophy but with rolling 5-hour windows rather than a monthly credit bucket. The era of unlimited AI subscriptions is ending....

I noticed a few interesting AI updates this week. 😎Claude Opus 4.7 was released. I haven’t tried it yet, but the first feedback I heard was not very excited. I’m not too motivated to test it, but anyway I’ll test it before making any conclusions. 😎GPT also has a new version - 5.5 This one looks promising, so I’m going to test it over the weekend. 🥲And GitHub Copilot changed its pricing strategy for individual plans. Now it looks closer to other coding agent providers: more limits, more structure, more usage control. Not the most exciting update. But probably a very expected one. What I liked is how GitHub explained the change. They were direct about the cost of agentic workflows and premium models. https://lnkd.in/diKTAf6P And I think this is the main point here: AI tools are growing fast, but the “unlimited usage” phase is slowly disappearing. For teams and engineers, this is worth watching. Not dramatic right now. Just another sign that AI coding tools are becoming more mature, more controlled, and probably more expensive 💰💰💰 Will test more and share thoughts later. #AI #GitHubCopilot #ClaudeCode #Codex #CodingAgents #EngineeringLeadership