Issue Station
Sign In
Tools hereToolbox

claude-code - 💡(How to fix) Fix [Bug] Claude Opus ignoring documented project context for Docker configuration [1 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
anthropics/claude-code#54207•Fetched 2026-04-29 06:33:26
View on GitHub
Comments
0
Participants
1
Timeline
5
Reactions
0
Author
kubi-devkubi-dev
Participants
kubi-devkubi-dev
Timeline (top)
labeled ×4subscribed ×1

Root Cause

For the last 2–3 days, the same model on the same projects suddenly bypasses all of that. It tries to
run bin/rails, bundle, rspec directly on the host. Of course it fails — I'm on macOS with system Ruby 2.6, the project needs 3.3, gems aren't installed locally, and that's exactly the reason the project is dockerized in the first place. The model then flails: tries to switch Ruby versions, suggests gem install bundler, asks me about rbenv. None of which is necessary because the answer is in the file it was supposed to read at session start.

RAW_BUFFERClick to expand / collapse

Bug Description Subject: Opus regression — model ignoring documented project context (again)

Hey Anthropic,

I have to flag what looks like another quality regression on Opus.

I run dockerized Rails applications. The setup is documented in CLAUDE.md / AGENTS.md / README files at the project root. For weeks, models read those files, picked up that everything runs through Docker, and used docker compose run / docker compose exec without issue. It just worked.

For the last 2–3 days, the same model on the same projects suddenly bypasses all of that. It tries to
run bin/rails, bundle, rspec directly on the host. Of course it fails — I'm on macOS with system Ruby 2.6, the project needs 3.3, gems aren't installed locally, and that's exactly the reason the project is dockerized in the first place. The model then flails: tries to switch Ruby versions, suggests gem install bundler, asks me about rbenv. None of which is necessary because the answer is in the file it was supposed to read at session start.

This is not a prompting problem on my end. The instructions haven't changed. The repo hasn't changed.
The only thing that changed is whatever you shipped on your side.

So a few honest questions:

  1. Did you silently nerf Opus again? Quantization, routing, "efficiency improvements," whatever you
    want to call it?
  2. If yes — when's the post-mortem? Because last time we got one a month later, after thousands of
    users had already eaten the degraded output and burned tokens debugging your regression instead of
    their code.
  3. If no — what changed? Because something clearly did, and "it's just vibes" is not a satisfying
    answer when I'm paying Max-tier prices.

Stop shipping silent quality changes to the most expensive model in the lineup and pretending nothing
happened until people complain loud enough on Twitter. If a model gets cheaper to serve at the cost of quality, say so. Give us a switch. Don't quietly downgrade the thing we're paying premium for.

Frustrated,

Environment Info

  • Platform: darwin
  • Terminal: vscode
  • Version: 2.1.119
  • Feedback ID: b907d92f-4a7a-418b-a452-e40b860f38d7

extent analysis

TL;DR

The Opus model may have silently regressed, causing it to ignore the documented project context and attempt to run commands directly on the host instead of using Docker.

Guidance

  • Review the CLAUDE.md, AGENTS.md, and README files at the project root to ensure they still contain the correct instructions for running the application through Docker.
  • Verify that the Docker setup and configuration have not changed, as the model's behavior suggests it is no longer reading the project context correctly.
  • Check the Opus model's version history to see if any recent updates or "efficiency improvements" may have caused the regression.
  • Consider reaching out to Anthropic support for more information on potential changes to the Opus model and to request a post-mortem analysis if a regression is confirmed.

Notes

The issue seems to be related to a change on Anthropic's side, but without more information, it's difficult to provide a definitive solution. The user's frustration suggests that this is not an isolated incident, and Anthropic's response will be crucial in resolving the issue.

Recommendation

Apply workaround: Until Anthropic confirms and addresses the potential regression, consider using a previous version of the Opus model or exploring alternative solutions that do not rely on the model reading the project context.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#installation#tensor shape#autograd error#model save/load#optimization

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING