Microsoft's latest shift toward a usage-based credit model for GitHub Copilot is sparking significant backlash among professional developers. Subscribers are reporting that their monthly credit quotas are draining at an alarming rate, with some users exhausting nearly 10% of their monthly allocation in just two hours of work. The change, initially reported by The Register, targets the more computationally expensive agental workflows that Microsoft claims are necessary for handling complex programming tasks.
The frustration isn't just about the cost, but the perceived value of the output. While testing tools like Claude 3.5 within the Copilot interface, developers have expressed disappointment with the results:
- Users have burned through over 1,100 credits in single sessions for site debugging.
- The quality of AI suggestions from integrated models has been described by some as mediocre relative to the credit cost.
- Long-term subscribers are now threatening to abandon the platform for alternatives that offer more predictable pricing.
Microsoft defends the implementation by highlighting that current AI models now execute far more resource-intensive logic than previous iterations. However, for many developers, the unpredictability of a credit-based system creates a barrier to the seamless coding experience Copilot originally promised.
