Michał Wesołowski’s Post

The latest changes to GitHub Copilot don’t just feel like a product update — they feel like a reality check. Tightened usage limits, reduced model access, and paused new sign-ups all point to one thing: the current model for AI developer tools may not be sustainable at scale. Yes, agentic workflows are more demanding. Yes, infrastructure costs are real. But from a user perspective, this shifts the burden back to developers — forcing us to think about tokens, limits, and model multipliers while trying to stay productive. That’s a step in the wrong direction. One of the biggest promises of tools like Copilot was to reduce cognitive load, not introduce a new layer of resource management. What’s more concerning is the signal this sends: If even a flagship product like Copilot needs to pull back on availability and tighten limits, what does that mean for the long-term viability of AI-assisted development as we know it today? Transparency improvements (like usage visibility in VS Code) are welcome — but they don’t address the core issue: 👉 The gap between how these tools are marketed vs how they actually scale under real-world usage. This feels less like a temporary adjustment, and more like the beginning of a pricing and capability correction across the industry. https://lnkd.in/gZuuVBE8 #GitHubCopilot

Stop Wasting Tokens: The 2026 GitHub Copilot Power Guide 🚀🛠️ Over the past few years, GitHub Copilot has evolved far beyond autocomplete. What used to be helpful suggestions is now closer to a system of specialized AI agents that can assist across your entire workflow. And with that shift, how we use it as developers is changing too. 🛠️ From prompting → to delegation Instead of relying on a single “do everything” approach, Copilot works best when you guide it clearly: • @terminal → for CLI, scripts, debugging • @docs → for accurate framework references • @test → for generating unit tests quickly 👉 Small shift, big impact on productivity ⚡ Thinking in systems, not steps One of the biggest unlocks is using tools like Composer for multi-file workflows. Instead of breaking tasks into many prompts, you can describe the outcome: “Add a Stripe webhook with a success email flow” …and let Copilot handle structure across files. 👉 Less back-and-forth, more momentum 🧠 Context matters more than ever Copilot performs best when the context is clear and focused. A few habits that help: • Keep only relevant files open • Use explicit references like #file:UserController.ts • Avoid vague descriptions when you can be precise 👉 Better context → better results 🧬 Let your types do the talking Providing structure (TypeScript interfaces, schemas) often works better than long explanations. It helps Copilot align with your system faster and more accurately. 🔁 Consistency improves results Using a simple structure for prompts: [Task] [Context] [Constraints] [Output Format] …can noticeably improve both output quality and efficiency over time. 🚀 The bigger shift As developers, the value is gradually moving from: Writing every line of code → Designing how systems get built Copilot is no longer just a tool you use. It’s something you collaborate with and guide. Curious how others are adapting their workflows—what’s been your biggest unlock so far? #GitHubCopilot #AIEngineering #SoftwareDevelopment #DeveloperProductivity #DevTools #GenerativeAI #TechLeadership #SeniorDevelopers #AIWorkflow

Our team uses GitHub Copilot with AGENTS.md files in each repository to give the AI context about our projects. Over time, we noticed the same dependency upgrade patterns being copy-pasted across multiple repositories: Framework migration steps Library compatibility matrices Known breaking changes and their fixes CI configuration patterns Instead of maintaining the same knowledge in 6+ places, we consolidated it into a single GitHub Copilot Agent Skill — a structured knowledge file that Copilot loads on demand when you need it. The result: 1,136 lines removed from scattered documentation files One source of truth, updated as we learn Today: the skill diagnosed a failing dependency upgrade in minutes — it already knew the root cause and the exact fix from the last time we solved a similar problem The real win isn't the line count. It's that next time someone on the team hits a dependency upgrade failure, the AI assistant already knows the solution from the last time we solved it. Knowledge that used to live in someone's head now lives in the toolchain. If you're using GitHub Copilot with AGENTS.md files, Copilot Agent Skills are worth looking into. Curious if others have found similar patterns for sharing AI context within a team. #GitHubCopilot #DeveloperExperience #DevOps #KnowledgeManagement

Hosted GitHub Copilot Dev Days – Exploring the Future of AI-Powered Development I recently had the privilege of hosting GitHub Copilot Dev Days, where I got to engage with an amazing community of developers and dive deep into how AI is transforming the way we build software. During the session, I covered some powerful aspects of the GitHub Copilot ecosystem: 🔹 GitHub Copilot in VS Code We explored how Copilot seamlessly integrates into VS Code to provide real-time code suggestions, improve productivity, and help developers write cleaner and faster code. Live demos showcased how it assists across different languages and use cases. 🔹 GitHub Copilot CLI One of the most exciting parts was demonstrating the Copilot CLI, where developers can use natural language directly in the terminal to execute commands, generate scripts, and simplify complex workflows. It truly brings AI assistance beyond the editor. 🔹 GitHub Copilot Cloud Agent We also discussed the Copilot Cloud Agent and how it enhances collaboration by enabling intelligent code generation, context awareness, and scalable AI support across projects and teams. The session was highly interactive, with great discussions around real-world applications, best practices, and how developers can effectively integrate AI into their daily workflows. It was inspiring to see the enthusiasm and curiosity among participants as they explored the potential of AI-assisted development. It was nice hosting this session and contributing to empower the developer community. Looking forward to many more such learning experiences! #GitHubCopilot #DevDays #AI #SoftwareDevelopment #VSCode #Innovation #DeveloperCommunity

GitHub is implementing changes to its Copilot Individual plans to ensure a more reliable and predictable experience for current users. I found it interesting that this shift aims to enhance user satisfaction amidst a growing reliance on AI tools. How do these changes align with your expectations or experiences using GitHub Copilot?

𝗚𝗶𝘁𝗛𝘂𝗯 𝗖𝗼𝗽𝗶𝗹𝗼𝘁'𝘀 𝗦𝗵𝗶𝗳𝘁 𝘁𝗼 𝗨𝘀𝗮𝗴𝗲-𝗕𝗮𝘀𝗲𝗱 𝗕𝗶𝗹𝗹𝗶𝗻𝗴 𝗙𝗲𝗲𝗹𝘀 𝗟𝗶𝗸𝗲 𝗮 𝗦𝘁𝗲𝗽 𝗕𝗮𝗰𝗸 Over the past months, I've run many GitHub & and GitHub Copilot workshops with engineering teams and enterprise stakeholders. One message came back again and again: 𝗚𝗶𝘁𝗛𝘂𝗯 𝗖𝗼𝗽𝗶𝗹𝗼𝘁 𝗴𝗮𝘃𝗲 𝘂𝘀 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆. Not just assistance. Not just governance. 𝗣𝗿𝗲𝗱𝗶𝗰𝘁𝗮𝗯𝗹𝗲 𝗰𝗼𝘀𝘁𝘀. With a huge margin, the most important argument organizations made in favor of Copilot adoption was 𝗯𝘂𝗱𝗴𝗲𝘁 𝗽𝗿𝗲𝗱𝗶𝗰𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆. Teams felt confident rolling Copilot out broadly because they could plan costs reliably. That was Copilot's real differentiator compared to most other AI coding tools. Switching to token-style billing (via AI Credits) changes that completely. If I compare this to platforms like OpenRouter, where I simply pay for what I actually consume, the prepaid credit approach feels like an unnecessary layer of abstraction rather than an improvement. Curious to hear how others in enterprise environments are reacting to this shift. Original News: GitHub Copilot is moving to usage-based billing https://lnkd.in/dcf4HJX6 #GitHub #Copilot

5 months ago I sat down with Scott Durow🌈 and Rami Mounla after Power Platform Community Conference to record where coding agents were heading. Rewatched it this week. Holds up better than most of what we were all saying back then. Governance muscle from the low-code years carries over. Who gets access. Which tools they can call. AuthZ/AuthN. How it gets reviewed when it goes to prod. One citizen dev with a coding agent just creates a lot more code in one run. The AI harnesses matured. Most of the practical progress went into instrumentation around the context window. Auto-compaction, handling context rot, steering through progressive tool descriptions. Copilot CLI made big jumps on that front. Primitives underneath are becoming standards - agent definitions, skills, hooks, MCP tools. What changed is how well the harness manages the window around all of that. My own opinion shifted too. In autumn the results were mostly there but I wasn't confident about it yet and I was saying so. Now it's not a question for me whether we go this way. Only who, how fast and how well. Coding agents run end-to-end through #UDPP26 next week. Scott opens with GitHub Copilot CLI tips for people who haven't started yet. We'll walk through the new CLI tools and skills Microsoft just released and where they fit with Power Platform. Then the full build lifecycle with agents. 🦸Diana Birkelbach on frontend. Jonas Rapp on backend. Matěj Samler on whole Power Platform solutions end-to-end. Raphael POTHIN on securing what those agents produce. Julie Koťátková on agents running UI tests against the result. Rami Mounla closes with the authoring and review workflow that keeps control of what actually ships. Governance track runs in parallel. Jan Hajek and Sabin Nair on Entra Agent ID and MCP in the tenant. Jukka Niiranen on inventory and cost. Marcel Ferreira and Casey Burke on ALM. 📅 April 27-28, Livestream + Q&A + recordings on demand + on-site if you're close #PowerApps #PowerPlatform #GitHubCopilot #GitHubCopilotCLI #CopilotStudio #Governance #ProDev #Dataverse

🚨 If you use GitHub Copilot, read this before June 1. I have suggested and evangelised that people should look at CoPilot Plus subscriptions to help them with their work, and help to plug gaps in their knowledge. Only fair that I warn every one of this upcoming change. GitHub just announced Copilot is moving to usage-based billing. From June 1, every plan Free, Pro, Business, Enterprise switches from flat-rate premium requests to token-based "AI Credits." What that means in plain English: → Every line of code AI generates for you now costs real, metered money → The fallback to cheaper models when you hit your limit? Gone → Agentic coding sessions that used to be "included"? Now billed per token → Your subscription price stays the same, but your actual spend won't This affects everyone: SME founders, solo developers, teams of any size, and anyone using Copilot to learn to code. The free AI coding era is officially over. But here's the thing I actually think this is a good thing. And I've written about why. Because this change is going to force a conversation most developers have been avoiding: do you actually understand what your AI-generated code does? Or does it just "look like what we want"? 👉 Full article: https://lnkd.in/eiGVXRw2 #GitHubCopilot #AI #SoftwareEngineering #SME #AIStrategy

🚨 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....