I've been saying this for two years. Get people hooked on AI tools. Make them dependent. Then change the pricing. GitHub just sent the email. Copilot is moving to usage-based billing on June 1. Your $10/month Pro plan now gets you $10 in "AI Credits." Your $39 Pro+ plan gets you $39. Sounds fair, right? Here's the problem: agentic sessions burn through tokens at a completely different rate than autocomplete. A single multi-step coding session can cost what used to feel like "unlimited." They're not wrong that the product changed. It genuinely did. Copilot today is not Copilot 2023. The compute costs are real. But the timing is also not a coincidence. You built your workflow around it. Your team built their workflow around it. Now the meter is running. This is not a GitHub problem specifically. This is the business model for every AI tool that gave you generous free or flat-rate access to establish adoption. If you haven't already: audit which AI tools your team actually uses and which ones you just have open out of habit. Because the bill is about to make that very clear. #AI #GitHub #Copilot #SoftwareDevelopment #TechLeadership
From what I'm seeing and hearing, not many companies have integrated AI into their workflows fully. Lots of companies have given their teams AI, and left them to figure it out. Therefore I think that the move away from unlimited usage at this time is less about trapping customers, and more about slowing their burn rate.
Ben Myrgorod
2d
And coding agents are used by people making 150-400k a year, right? They are just becoming expensive professional tools. Companies will afford that. Days you can run LLM systems like that locally are coming soon so no reasons to worry.
I have been trying opencode with ollama cloud API. I am surprised at how close it is matching other setups now.
Ian Morgan
4d
And this is why local AI, knowledge of the models and the technology stack, and access to safe model repositories like Microsoft Foundry should be part of your AI strategy going forwards.
This is real. I use GitHub Copilot and Claude Code daily. Before, it felt unlimited. You did not think about cost. Now every prompt has a price. In real work, agents burn fast. One session can eat what used to last all day. So the habit has to change. I keep prompts tight. I break tasks down. I avoid long runs. AI is still a win ✅ But now it needs control, not just convenience.