Petter Arnesen’s Post

GitHub Copilot app is launched in tech preview!! 🎉 TL;DR: The GitHub Copilot app gives devs and dev-adjacent users a native desktop home for managing agentic work across issues, PRs, reviews, checks, merges, repeatable workflows and life in general. I'm obviously no dev, so the VS plugin is a little much for me. But this is 🔥🔥🔥

Testing the new GitHub Copilot desktop app over the past few days has been promising. It feels like a meaningful step in the right direction. Skills, MCP servers, workflows (resembling Codex automations), and an intuitive chat experience are beginning to converge into a more capable developer environment. For those familiar with Codex, the experience will feel immediately familiar. What stands out however is the enterprise angle. For many customer segments, remaining within an enterprise governed ecosystem is not optional. It is essential. This creates a double edged dynamic for both customers and vendors. Yet also a critical balance neither can afford to ignore. This tension may create adoption challenges for frontier coding agents from OpenAI or Anthropic regardless of capability. GitHub Copilot’s trajectory therefore matters. Particularly in the near term. While this is a promising development, execution will ultimately determine the trajectory. Reliability and consistency of the GitHub experience remain critical. Enterprise licensing and commercial models need to become far more streamlined. And rapid iteration of the Copilot App will be essential to sustain engagement and establish credibility alongside platforms like Codex and Claude. Reliability, enterprise licensing, and iteration speed must emerge as defining priorities for GitHub Copilot’s next phase. The coding agents race may not be won by capability alone. In enterprise environments, trust, governance, distribution, and execution velocity may matter just as much. What matters more: Capability or enterprise readiness? #AI #GitHubCopilot

GitHub Copilot just got 900 downvotes for being honest. They moved all paid users to usage-based billing this week. Same monthly price, but now it converts to credits. Run out, and premium features stop until next cycle. Developers are furious. The comments are brutal. But here's what nobody's saying: this was always what AI tools were going to cost. The flat subscription era was a land grab. Every major AI company priced below cost to get you hooked. GitHub was no different. Usage-based billing isn't a betrayal. It's just honesty. The real lesson for SMB owners isn't "avoid GitHub Copilot." It's: before you build your business around any AI tool, understand what it actually costs at scale. Cheap during the trial period. Metered once you depend on it. Expensive once you've built workflows around it and switching hurts. This is exactly why strategy comes before tools. Not the other way around. If you're evaluating AI for your business right now, understand the pricing model before you fall in love with the features. What AI tools are you paying for that you haven't stress-tested at real volume? #ai #smallbusiness #aitools #aiautomation #github

GitHub Copilot goes pay-per-token June 1. Your $10/month Pro plan now buys exactly $10 in AI Credits. One agentic coding session could burn all of it. GitHub is killing flat-rate Copilot pricing this weekend. Everything except code completions and Next Edit Suggestions moves to usage-based billing measured in AI Credits worth $0.01 each. The per-seat price stays the same, but what that price buys is now metered. The math is ugly. A community developer estimated agentic coding sessions routinely consume $30 to $40 per session. Copilot Pro includes $10 in monthly credits. You could blow through your entire month in a single afternoon. Free models are gone. GitHub's exact words: "free models are no longer part of our offering." Annual plans are being retired too. When your annual sub expires, you drop to Copilot Free unless you re-up monthly. GitHub's argument is that Copilot "now powers far more complex, agentic workflows that consume far more compute." Fair point. But they built those features, marketed those features, and now they're charging you for using what they sold you. I track AI spend across every client project in my agency. Here's what I'm telling people this week: - Open the billing preview in your editor or GitHub.com before June 1. See what your usage actually looks like in credits. - Set your overage cap now. Don't discover a $200 weekend by accident. - Code completions are still unlimited. If that's most of your Copilot usage, you might be fine. - If you rely on agentic workflows, compare direct API access. You may get more mileage paying the model provider directly instead of through GitHub's layer. The $10/month unlimited AI coding era ends Sunday. Every platform will follow this model. The question isn't whether usage-based billing is fair. It's whether your budget is ready. #GitHub #Copilot #AIBilling #DeveloperTools #UsagePricing

GitHub Copilot's new billing system starts TODAY (June 1). WHAT'S CHANGING: → Premium Request Units → GitHub AI Credits → Credits consumed based on token usage → Once credits run out, buy more or wait PLAN PRICES (unchanged): - Pro: $10/month = $10 in credits - Pro+: $39/month = $39 in credits - Business: $19/user = $19 in credits - Enterprise: $39/user = $39 in credits WHAT'S FREE: ✓ Code completions ✓ Next Edit suggestions WHAT'S GONE: ✗ Fallback to lower-cost models when credits are exhausted PROMOTIONAL CREDITS (June-Aug): - Business: $30 instead of $19 - Enterprise: $70 instead of $39 DEVELOPER REACTION (TechCrunch): → Some projecting $29 → $750/month increases → Others saying $50 → $3,000 → Users switching to Claude, DeepSeek, and Codex GitHub CPO Mario Rodriguez: "The current premium request model is no longer sustainable." Critics: "This is completely unacceptable." Defenders: "The people spending this much are vibe-coders." What's your take? https://lnkd.in/gYVjKsNi #GitHubCopilot #AI #Developers #Pricing

GitHub Copilot is officially DEAD GitHub Copilot went from a $10 tool to a potential $100+ habit real quick 💀 GitHub added a new pricing simulation tool to compare the old Copilot model vs the new usage-based billing system, and the numbers genuinely surprised me. Earlier it was simple: Pay ~$10-$20 and use it without thinking too much. Now the simulation shows: • 12M+ AIC usage • Estimated bill: $116+ Of course this is just a simulation, not an actual bill. But it clearly shows how expensive AI-assisted development can become once pricing shifts to pure usage. And honestly, this feels similar to cloud costs: Everything looks cheap initially… until scale kicks in. The worrying part is that developers may now start thinking before every heavy AI workflow: “Should I really use Copilot for this?” AI coding tools are becoming essential now, but predictable pricing mattered a lot more than we realized. Curious how other devs feel about this shift. Would usage-based pricing change how often you use AI coding assistants? Check yours estimated bill- https://lnkd.in/gk9W8ZYZ #GitHub #GitHubCopilot #AI #SoftwareEngineering #Developers #Programming #Tech #Coding

GitHub Copilot is changing its billing model on June 1st, and I think a lot of people might get a surprise. I checked the preview tool for my April usage. Separate from work stuff, I use Copilot Pro for personal projects, waaaay less than for consulting work. Under the current model my bill was $18.63. Under the new usage-based model, the same usage would cost $118.27. That's a 6x increase. For light personal use. Copilot is moving from fixed premium requests (where every interaction costs the same regardless of complexity) to token-based billing. "commit and push" used to cost the same as "analyse my codebase for simplification opportunities". Now they don't. GitHub absorbed the difference for years. For business users running Copilot in agentic workflows, the numbers could be even more stark. If you're managing a team budget, I'd strongly recommend checking the preview tool before June 1st. Links in the comments. For me personally, it's a nudge to get serious about local models. I run Ollama on an M1 Max Mac Studio with 64GB unified memory, which can handle 32B parameter models reasonably well. The question is whether that's good enough for real coding assistance. I'm going to find out. The era of AI services being underpriced to drive adoption is ending. Labs and platforms are aligning pricing with actual compute costs. That's not unreasonable, but it does change the calculus for personal projects, side hustles, and small teams who assumed these tools would stay cheap. If you use Copilot, check your numbers. The preview tool is live now. #GitHubCopilot #AI #MachineLearning #SoftwareDevelopment #AITools

GitHub Copilot just gave developers a nasty billing surprise. On 1 June, GitHub flipped the switch on token-based billing for Copilot. And some developers are staring at projected cost increases of up to 50x. The backlash has been immediate and, frankly, predictable. Developers hate billing surprises. Especially when the surprise arrives mid-sprint with no graceful transition. Here's the real issue: usage-based pricing for AI coding tools creates anxiety. You stop asking the tool questions. You second-guess every prompt. The cognitive overhead of watching a meter tick up actively undermines the productivity gains you bought the tool for in the first place. I wondered how long it would take before billing changed, it does feel a little like they have you hooked and so now increase the costs. Surely there is a sensible version of this model. Tiered caps. Clear dashboards. Predictable spend. Pricing models matter as much as product quality. Every AI vendor in this space should be paying close attention right now. What do you think? Nessesary move to create revenue increase? Sources: → https://lnkd.in/e4d_QkNM #GitHubCopilot #AITools #DeveloperExperience #SoftwareDevelopment #AIBilling #TechNews

🚨 GitHub Copilot Limits the "Unlimited": The Transition to Token-Based Pricing is Here! 🚨 If you’re a developer relying heavily on GitHub Copilot, it’s time to check your billing settings. Starting June 1st, GitHub has officially transitioned its pricing model from the traditional flat-rate subscription to a Token-usage Billing system. The era of unlimited, unrestricted requests under the $10 Pro Plan is officially over. 🛑 Here is a breakdown of what this massive shift means for developers and tech teams: The Shift to AI Credits: GitHub is phasing out Premium Request Units (PRUs) and introducing GitHub AI Credits. Under this new system, 1 AI Credit = $0.01. What remains Unlimited? Fortunately, standard Code Completions and Next Edit Suggestions remain unlimited for now. What will drain your credits? Heavy-lifting features like Copilot Chat, Agentic Workflows, and automated Code Reviews will now be billed dynamically based on your exact token usage. Estimated Plan Allocations: Pro Plan ($10/mo): Will roughly include 1,000 to 1,500 credits. Pro+ Plan: Will roughly include 7,000 credits. 📉 The Developer Community Reaction The dev community is already feeling the heat. Heavy users and enterprise developers are projecting that their previous $39/month workflows could skyrocket to as much as $300/month under the token system. Because of this massive cost spike, a noticeable wave of developers are already canceling their subscriptions. 💡 Pro-Tip: Set a Budget Limit Today! If you plan to continue using GitHub Copilot, my strongest advice is to log into your account immediately and set a strict Budget Limit (Hard Limit). If you don't cap your usage, your next credit card bill might come as a very unpleasant surprise! 💸 What are your thoughts on this new token-based pricing? Are you sticking with Copilot, or is this the perfect excuse to switch to alternatives like Cursor or Claude? Let's discuss in the comments! 👇 #GitHub #GitHubCopilot #ArtificialIntelligence #SoftwareEngineering #TechUpdates #Coding #Developers #DevCommunity #TechPricing

GitHub Copilot just switched from a flat monthly subscription to token-based billing. That is a bigger deal than most people are treating it as. Flat subscription meant you could use Copilot as much as you wanted for a fixed cost. Token-based billing means every autocomplete, every suggestion, every generation now has a meter running. The more you rely on it the more you pay. Two things this does immediately: First, it changes the psychology of how you use the tool. When something is unlimited you use it freely. When every use has a cost attached you start asking whether this particular suggestion is worth it. That question is actually useful. It forces you to stay in the loop rather than defaulting to whatever the model outputs. Second, it reveals what GitHub actually thinks Copilot is worth. Token pricing is what you do when you believe usage will grow dramatically. They are betting that developers will use significantly more AI assistance over time and they want a piece of every token of that growth. For students using the free GitHub education tier — check your limits. The pricing change may affect what is available to you. For everyone building AI-assisted tools right now — this is a reminder that building on top of someone else's API means their pricing decisions become your problem. Local inference is looking more attractive every week. #AI #GitHub #Copilot #BuildInPublic #AIEngineering