Data and AI struggles

1,730 problems · 342 companies

LLM apps

Prompts break or get ignored when models change

1 post from 1 person1d activeFading, +0 this weekunverified

Posts per day

Posts per day1 posts, Oct 27 to Nov 9

The posts behind it

GitHub issuesanthropics/claude-code · timoncoolOct 27

[MODEL] Opus 5.5 in Claude Code regressed since the week of Oct 19, unusable by Oct 26–27 (Max 20x) — and there is no way to reach a human at Anthropic

## Summary I am a paying Max 20x subscriber ($200 every month) and a heavy daily user of Claude Code. Over the last two weeks Claude Code with Opus 5.5 has regressed to the point where working with it is no longer possible: - it ignores the instructions, rules and skills it has...

13 reactions2 replies

Companies and products named

About this problem

Evidence

1 post from 1 person in 1 place. Mostly on anthropics/claude-code. Tools named alongside: Claude Code, Claude, Model Context Protocol, GitHub.

Frustration Frustration 3 of 3· Seen on GitHub issues

How it was grouped

Posts that state a pain and match the "prompt-brittle" rule. First post Oct 27, 2026, latest Oct 27, 2026. Unverified: fewer than 3 posts, or one person or place only. Method

History

Momentum went from new to fading.

Nov 3, 2026

First seen with 1 post.

Oct 31, 2026