How CRAFT Builds Ketelsen.ai — The Framework Behind the Prompts
Every week, Ketelsen.ai publishes prompt posts that work across ChatGPT, Claude, and Gemini — posts with structured breakdowns, real industry examples, and difficulty levels that scale from beginner to advanced. What most readers don't see is the system that makes that consistency possible. It's called CRAFT — Configurable Reusable AI Framework Technology — and it's the engine underneath everything you read on this site. This post pulls back the curtain — what CRAFT is, why it exists, how it powers every stage of the Ketelsen.ai weekly workflow, and why it's the reason Ketelsen.ai 2.0 can promise something most AI content sites can't: measurable improvement over time.
What Is CRAFT?
CRAFT (Configurable Reusable AI Framework Technology) is a structured system for communicating with AI that borrows ideas from object-oriented programming and applies them to AI conversations. It was developed by Richard Ketelsen at Ketelsen Digital Solutions LLC and has been evolving through real-world use since late 2025.
The core idea is straightforward: most people interact with AI as if every conversation is a blank page. No memory of what worked last time, no consistent format, no quality checks, no way to build on what you learned yesterday. CRAFT solves that by introducing the same kind of structure that software engineering uses to manage complexity — but adapted for natural language AI workflows.
In practical terms, CRAFT provides six capabilities that generic AI conversations don't have:
Recipes — reusable instruction sets that produce consistent results every time they run. Think of them as tested, version-controlled procedures for specific tasks. Ketelsen.ai currently uses over 150 recipes organized across multiple cookbooks.
Session Handoffs — a structured system that preserves context, decisions, and project state across conversations. Normal AI chats start from zero every time. CRAFT picks up exactly where the last session left off.
Lessons Learned — a formal system that captures what works and what doesn't, then makes those insights permanent. When we discover that a certain prompt structure produces better industry examples, that knowledge gets codified and applied to every future post.
Personas — role-based AI configurations that specialize in different aspects of the workflow. Ketelsen.ai uses a multi-persona system where different specialists handle content creation, quality validation, and operations management.
Version Control — every project file, recipe, and lesson is tracked in git. We can see exactly what changed, when, and why. If something breaks, we trace it back.
Quality Gates — automated structural validation that catches formatting issues, missing sections, and inconsistencies before anything reaches readers.
If you want to explore CRAFT in detail, visit craftframework.ai.
Why CRAFT Exists: The Problem It Solves
AI is powerful. But without structure, it's also unpredictable. Anyone who has used ChatGPT, Claude, or Gemini for real work has hit the same walls: the AI gives a great answer on Tuesday and a mediocre one on Wednesday with the same prompt. You can't remember what instructions produced last month's best output. You spend fifteen minutes re-explaining context that the AI should already know. Quality depends entirely on how much energy you have that day.
These aren't AI problems. They're workflow problems. And they're the exact problems that software engineering solved decades ago with version control, modular code, testing frameworks, and documentation standards. CRAFT takes those proven concepts and translates them into a system that works inside AI conversations.
The result is that content quality stops depending on luck or effort and starts depending on a system that improves every time it runs.
How CRAFT Powers Ketelsen.ai 2.0
Ketelsen.ai 1.0 was a manual operation. Research a topic, write the prompts, test them, publish. It worked, but every decision lived in one person's head. The quality of each week's output depended on available time, mental freshness, and whether the author remembered what worked three weeks ago. There was no system — just effort.
Ketelsen.ai 2.0 is what happens when you point a mature CRAFT implementation at that same workflow. Here's how CRAFT touches every stage of the weekly content pipeline:
Stage 1: Topic Selection (Research-Driven)
Instead of guessing what readers need, the 52-week content calendar was built from 17 deep research files covering common AI frustrations, prompt literacy gaps, underutilized features, and emerging capabilities. Every weekly topic maps to specific research findings. CRAFT's project file system keeps that research accessible across sessions — so Week 25 can reference the same research that informed Week 1 without re-uploading anything.
Stage 2: Prompt Generation (Multi-Platform, Multi-Level)
Each week's prompts are generated using a structured recipe that establishes context across AI platforms, then produces three variations — Beginner, Intermediate, and Advanced — that share a topic but scale in complexity. A standardized multi-section blog post template ensures every post follows the same structure. Before CRAFT, each post's format drifted slightly depending on when it was written. Now there's a single source of truth.
Stage 3: Cross-Platform Testing
Every prompt is run through ChatGPT, Claude, and Gemini. The results are scored using a 7-dimension quality rubric that measures prompt engineering depth, breakdown clarity, industry example specificity, writing quality, creative use cases, actionability, and completeness. This isn't a gut-feel comparison — it's a structured evaluation with weighted dimensions and anchor descriptions at every score level.
Stage 4: Quality Validation
Before anything is published, automated structural checks verify section completeness, formatting consistency, and template compliance. The most recent validation runs scored a perfect 10 out of 10. Quality gates are not optional extras — they're built into the workflow.
Stage 5: Continuous Improvement
After every session, CRAFT's lessons-learned system captures discoveries that improve future output. When we find that named creative angles produce more diverse results than open-ended instructions, that becomes a permanent improvement. When we discover that external template files prevent format drift better than inline specifications, that gets codified. Nothing gets lost between sessions.
The Difference at a Glance: Traditional AI vs. CRAFT
What Happens Without CRAFT
Between Sessions
Automatically
(If Any)
of AI Interactions
What Happens With CRAFT
via Handoffs
& Applied
(10/10 Recent Score)
& Recoverable
Chart 2: How CRAFT Flows Through the Ketelsen.ai Weekly Pipeline
17 Deep Files
52-Week Calendar
10-Part Prompt
3 Variations
ChatGPT + Claude
+ Gemini
7-Dim Rubric
Structural QA
HTML Assembly
Chart 3: CRAFT By The Numbers — Ketelsen.ai Project Stats
| Metric | Count | Why It Matters |
|---|---|---|
| Total Recipes | 159 | Reusable procedures covering session management, content creation, quality validation, git operations, and more |
| Active Cookbooks | 5 | Organized recipe collections: Core, Studio, Admin, Cowork, and Cowork Admin (spanning 4 specialized volumes) |
| Lessons Learned | 29 | Permanent insights captured from real sessions — each one prevents a mistake from happening again |
| Session Handoffs | 16+ | Seamless context transfers between sessions — Week 10 knows everything from Weeks 1 through 9 |
| Blog Post Template Sections | 13 | Standardized format ensures every post is structurally complete — no missing sections, no drift |
| Quality Rubric Dimensions | 7 | Weighted scoring across prompt quality, clarity, examples, writing, creativity, actionability, and completeness |
| Content Calendar Weeks | 52 | A full year of topics mapped to research — no guessing, no filler weeks |
| AI Platforms Tested Weekly | 3 | ChatGPT, Claude, and Gemini — same prompt, same rules, transparent comparison |
Chart 4: The Compounding Effect — How CRAFT Accumulates Intelligence Over Time
Each bar represents the cumulative knowledge base available to the system at that point. Traditional AI workflows (gray) reset to zero every session. CRAFT (orange) compounds.
Three CRAFT Principles You Can See in Every Post
CRAFT is built on six philosophical principles. Three of them directly shape what you experience as a Ketelsen.ai reader — and they're worth understanding because they explain why the posts work the way they do.
Structure Enables Freedom. By providing a clear framework for every AI interaction, CRAFT eliminates the cognitive overhead of re-inventing the process every week. That freed-up mental energy goes directly into creativity — better prompts, more unexpected use cases, sharper industry examples. When we're not spending time on "how do we format this post," we're spending time on "what's the most useful prompt we can build."
Cumulative Intelligence. Most AI conversations start from zero every time. CRAFT doesn't. Every session adds to a permanent record that future sessions draw from. When we discovered in Week 1 that named creative angles produce more diverse results than open-ended instructions, that insight became permanent — and it improved every post that followed.
Human-AI Partnership. CRAFT doesn't replace human judgment — it amplifies it. The human sets direction, makes editorial calls, and decides what "good" means. The AI handles scale, consistency, and structural execution. The result is output that neither could produce alone: human taste at AI speed, with a system that remembers everything both sides have learned.
The remaining three principles — Code as Conversation, Accessibility Through Familiarity, and Conservation of Tokens — are more relevant to people building their own CRAFT implementations. You can read about all six at craftframework.ai.
What This Means for You
If you're a Ketelsen.ai reader, here's the most important takeaway: the prompts are going to keep getting better. Not because we're working harder, but because the system that produces them is designed to improve itself. Each week's output builds on every lesson from every previous week. The difference between Week 1 and Week 52 won't be subtle — and CRAFT is the reason.
Every prompt on Ketelsen.ai continues to feature the same practical, creative approach you expect — clear instructions, real examples, techniques that work across all major AI platforms. The difference is that the system producing them is now more rigorous, more consistent, and more capable of surfacing ideas that wouldn't have occurred to a solo author on a deadline.
Want to see CRAFT in action? Read this week's prompt post and notice the structural consistency across all three variations — the same template, the same quality standards, the same depth — regardless of which AI platform produced it. That consistency is CRAFT at work.
And if you're curious about building your own CRAFT-powered workflow, visit craftframework.ai to learn more about the framework.
Published March 2026. Ketelsen.ai 2.0 is powered by the CRAFT Framework, developed by Ketelsen Digital Solutions LLC.
