The "Compaction" Update has rendered VS Code Agents non-functional #191955
Replies: 1 comment
-
|
💬 Your Product Feedback Has Been Submitted 🎉 Thank you for taking the time to share your insights with us! Your feedback is invaluable as we build a better GitHub experience for all our users. Here's what you can expect moving forward ⏩
Where to look to see what's shipping 👀
What you can do in the meantime 💻
As a member of the GitHub community, your participation is essential. While we can't promise that every suggestion will be implemented, we want to emphasize that your feedback is instrumental in guiding our decisions and priorities. Thank you once again for your contribution to making GitHub even better! We're grateful for your ongoing support and collaboration in shaping the future of our platform. ⭐ |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
-
🏷️ Discussion Type
Product Feedback
💬 Feature/Topic Area
Copilot in GitHub
Body
I am writing this because the recent changes to context compaction have effectively lobotomized the VS Code agents. What used to be a semi-reliable pair programmer has been reduced to a single-turn task runner with the memory of a goldfish.
The current state of "compaction" is a disaster for any real-world workflow. After just a few exchanges, the agent:
Goes totally blind: It loses track of the very file structures and variable names we were discussing two minutes prior.
Becomes directionless: It forgets the overarching goal of the session, requiring me to manually re-explain the entire context every three prompts.
Hallucinates legacy state: Because the "compacted" summary is so vague, the agent starts guessing and inventing logic that doesn't exist in the current codebase.
This isn't "optimization"—it's a massive regression. We are paying for "agents," not "single-task bots." If the goal was to save on token costs by sacrificing the entire utility of the tool, you’ve succeeded.
Is there a way to opt-out of this aggressive summarization, or do we just have to wait for a competitor to provide an agent that can actually remember a conversation?
Beta Was this translation helpful? Give feedback.
All reactions