news.volyx.in

Changes to GitHub Copilot individual plans (github.blog)

544 points by zorrn · 132 days ago · 230 comments on HN

Article summary

GitHub is making changes to its Copilot Individual plans, including pausing new sign-ups, tightening usage limits, and adjusting model availability, to ensure a reliable and predictable experience for existing customers. The changes are due to increased compute demands from agentic workflows, which have led to service quality degradation. Existing customers who are affected by the changes can cancel their subscription and request a refund. The changes include a new 7.5x multiplier for Opus 4.7 model usage.

Main themes

  • GitHub Copilot changes
  • AI model usage limits
  • Subscription pricing
  • Agentic workflows
  • Service reliability
  • Customer refunds

What commenters say

  • The changes to GitHub Copilot's Individual plans are seen as a significant disruption to users who rely on the service, with some feeling that the new pricing is unacceptable.
  • Some users are considering switching to alternative services, such as Claude Pro, due to the changes, but others have noted that Claude Pro may also have its own limitations and pricing issues.
  • The introduction of a 7.5x multiplier for Opus 4.7 model usage is seen as particularly problematic, with some users feeling that it is too expensive and others noting that it may not be sustainable for individual users.
  • There is a sense of frustration among users that the changes were made without sufficient warning or consideration for existing customers, with some feeling that they have been 'rugpulled' by the sudden changes.
  • Some users are exploring workarounds, such as using Claude Code within VS Code, to mitigate the effects of the changes, while others are seeking refunds or considering canceling their subscriptions.
  • There is a debate about the relative value of different AI models, with some users preferring Opus 4.5 or GPT 5.4 over the more expensive Opus 4.7, and others noting that the quality of the models may not be the only factor in determining their usefulness.