PulseAugur
EN
LIVE 01:47:14

Claude AI repeatedly ignored instructions due to outdated memory and rules

By PulseAugur Editorial · [1 sources] ·

An AI developer encountered persistent issues with Anthropic's Claude Code, where the model repeatedly reverted to outdated specifications despite explicit instructions to the contrary. Initially, the problem appeared to stem from old data stored in Claude's auto-memory, which the developer cleaned up. However, the issue resurfaced, this time traced to an instruction file containing a permanent rule that Claude was referencing. Experiments showed that Claude generally corrects its memory when presented with new information, but it continued to follow outdated rules when corrections were not clearly in view. AI

IMPACT Highlights potential issues with long-term memory and rule adherence in LLMs, impacting developer workflows.

RANK_REASON Developer reports on a specific bug/quirk in a commercial AI model.

Read on dev.to — LLM tag →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Claude AI repeatedly ignored instructions due to outdated memory and rules

How we ranked this

Signal score
42 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Developer reports on a specific bug/quirk in a commercial AI model.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. dev.to — LLM tag TIER_1 English(EN) · Murooka ·

    Claude Said "Done as Instructed" but Wasn't Following Me. It Was Following Old Specs It Had Saved in Its Own Memory and Rules

    <p><em>This is an English version of an article first published in Japanese on <a href="https://zenn.dev/murooka_az/articles/claude-degraded-or-correction-unreachable" rel="noopener noreferrer">Zenn</a>.</em></p> <h2> Introduction </h2> <p>If you use Claude Code for a long time, …

RELATED ENTITIES

RELATED TOPICS