# WebMnem AI Intelligence

> Premium mobile-first newsroom and local-agent operations lane for WebMnem AI.

Canonical page: https://webmnemai.netlify.app

## What this site covers

- WebMnem AI front-end/editorial UX systems
- canonical article routes for modal stories
- local-agent harness design
- MLX-LM and Ollama local model workflows
- prompt caching, LoRA, evals, receipts, and state hashes
- here.now publishing discipline
- mobile-first newsfeed and article-reader patterns
- applicant proof surfaces and role-archetype WebMNEM routes

## Primary proof routes

- AI Workflow Infrastructure Architect: https://webmnemai.netlify.appai-workflow-infrastructure-architect/
- AI Workflow Specialist: https://webmnemai.netlify.appai-workflow-specialist/

- AI Evaluation Workflow Assurance proof surface: https://webmnemai.netlify.appai-evaluation-workflow-assurance/
## Current doctrine articles
- Same Model, Different Operator: Prompt Fencing as Runtime Control: https://webmnemai.netlify.appsame-model-different-operator-prompt-fencing-runtime-control/
  - Summary: Prompt fencing turns prompting into runtime control: lock objectives, forbid substitutions, preserve partials, and prove completion before an agent can call work done.

- OKF and MCP: Build the AI Context Stack Before the Chatbot: https://webmnemai.netlify.appokf-mcp-ai-context-stack/
  - Summary: Before attaching a chatbot to business systems, separate durable organizational knowledge from governed live tool access.

- Don’t Use a Stochastic Engine for Deterministic Workflows: https://webmnemai.netlify.appdont-use-a-stochastic-engine-for-deterministic-workflows/

- Markdown Doesn't Compel Generation: https://webmnemai.netlify.appmarkdown-doesnt-compel-generation/
  - Summary: Markdown documents intent, but generated artifacts need executable contracts, browser behavior tests, answer checks, receipts, hashes, and regression gates.

## Authority note

The website is a public editorial artifact. Canonical implementation state lives in the local WebMnem AI repository through source files, receipts, docs, and state hashes. Local-agent output is advisory until verified by a supervising agent.