Skip to content

Latest commit

 

History

History
68 lines (54 loc) · 2.78 KB

File metadata and controls

68 lines (54 loc) · 2.78 KB
name claude-intel-monitor
description Detect when Claude/GPT/DeepSeek gets dumber. 30 benchmark questions (Math/Reasoning/Code), response quality analysis, ZenMux smart routing, model switch recommendations. Track intelligence degradation over time.
version 1.2.0
author minirr890112-byte
license MIT
metadata
hermes
tags homepage
AI
Monitoring
Benchmark
Claude
GPT
DeepSeek
Intelligence
Degradation

Claude Intel Monitor

Problem → Solution

The problem: You're coding with Claude and suddenly it feels... off. Code completions are simpler. Reasoning skips steps. Math answers are wrong. You wonder: "Did they nerf the model?" Anthropic and OpenAI don't announce degradations. You're left guessing if it's you or them.

The solution: Standardized 30-question benchmark across Math, Reasoning, and Code. Run it daily, track scores over time. When scores drop >10%, you get an alert. Response quality analysis detects thinking-skipping. ZenMux routes to backup models when your primary degrades.

Quick Start

pip install git+https://github.com/minirr890112-byte/claude-intel-monitor.git

claude-intel-monitor test --model claude-sonnet-4 --provider anthropic
claude-intel-monitor watch --model claude-sonnet-4 --interval 6h
claude-intel-monitor quality -f response.txt -c code
claude-intel-monitor history
claude-intel-monitor zenmux
claude-intel-monitor switch -s degraded

Real Output

$ claude-intel-monitor test --model claude-sonnet-4

🧠 Claude Intel Monitor v1.1.0
Model: claude-sonnet-4 (anthropic)
────────────────────────────────────
Math:      8/10 (80%) ████████░░
Reasoning: 7/10 (70%) ███████░░░
Code:      9/10 (90%) █████████░
────────────────────────────────────
Total:     24/30 (80%)
Status: ✅ Normal (within 5% of baseline)

$ claude-intel-monitor quality -f response.txt
Response Quality: 0.62/1.00 ⚠️
├─ Thinking depth:   0.45  ⚠️ Shallow
├─ Code completeness: 0.85  ✅
└─ Hallucination:    0.55  ⚠️ Suspect

Commands

Command Description v1.1
test Run 30-question benchmark
watch Continuous monitoring at interval
quality Response quality analysis
zenmux Smart routing to backup models
switch Model switch recommendations
history Score trend over time

Star this repo if you've felt the model get dumber: github.com/minirr890112-byte/claude-intel-monitor