From zero AI vocabulary to Shela-level working habits: clear prompts, CLI fluency, evidence-based research, careful publishing, and repeatable systems using Codex CLI and Claude CLI.
A beginner does not need to start with the terminal. The first win should match how they naturally communicate: chat, search, file work, code work, or research. Pick one method, finish one small proof, then move into the full blueprint.
Best first useThinking partner for messy ideas, lesson notes, self-reflection, long explanations, and critique before writing.
First winTurn one vague goal into a clear learning plan with outcome, constraints, next action, and proof.
Suggested lessonsTake lessons 2, 5, 6, 7, and 23.
Starter message: I am new to AI. Help me turn this messy goal into a clear task, then ask me what source material you need before giving advice.Best first useGuided codebase reading, file-aware planning, documentation cleanup, and careful editing when the work lives inside a project folder.
First winAsk it to read a small project, explain the structure, and suggest one safe improvement without editing yet.
Suggested lessonsTake lessons 9, 12, 15, 16, and 23.
Starter message: Read this folder first. Explain what each important file does, then suggest one small improvement and how we would verify it.Best first useFast practice partner for beginner questions, drafts, roleplay, checklists, simpler rewrites, and confidence reps.
First winCreate one reusable prompt recipe and test it on a small real task.
Suggested lessonsTake lessons 1, 5, 8, 13, and 24.
Starter message: Teach me this like a beginner. Give me a simple example, then make me practice with one small task and check my answer.Best first useImplementation partner for files, scripts, HTML pages, tests, verification, and local tool building.
First winBuild or edit one tiny local artifact, then inspect the diff and run the verifier before calling it done.
Suggested lessonsTake lessons 9, 10, 12, 15, 16, 21, and 22.
Starter message: Inspect the current files first. Make the smallest useful change, run the relevant check, and tell me exactly what evidence proves it works.Best first useBroad brainstorming, Google-centered work habits, comparing options, and quick explanations when the learner is still exploring.
First winCompare three possible learning routes and choose one based on time, tools, and proof.
Suggested lessonsTake lessons 1, 2, 5, 17, and 19.
Starter message: Compare these options for a beginner. Tell me which one is easiest to finish this week, what proof I should create, and what to avoid.Best first useSearch-first learning, current-topic exploration, source discovery, and follow-up questions that need web context.
First winResearch one unfamiliar AI term, collect useful links, and turn them into a simple source-backed note.
Suggested lessonsTake lessons 14, 17, 18, 19, and 20.
Starter message: Help me research this topic from current sources. Give me the useful links, explain what each source proves, and separate facts from assumptions.The map is strong because it teaches judgment, not tricks. The danger is that different beginners will get overwhelmed in different places. These notes show where each archetype should enter, what they will resist, and what proof they should build first.
What worksThis learner will understand the production loop quickly. They already know client pressure, deadlines, and the need for proof.
RiskThey may skip Orientation and Prompt Craft because they feel experienced. That would leave them using AI as a faster assistant instead of a system partner.
First methodStart with Codex if the proof is a tool or page. Start with Claude Web if the proof is a case study or client process.
First pathStart at lessons 13 to 16, then 25 to 32. Their goal is to turn existing freelance proof into verified AI-enabled workflows.
What worksThe guardrails section will feel familiar because corporate and BPO work already trains people to follow process, escalate issues, and document outcomes.
RiskThe CLI section may feel intimidating. They may think terminal work is only for programmers, not for operations people.
First methodStart with ChatGPT Web for confidence reps, then Claude Web for translating work history into remote-work tasks.
First pathStart at lessons 1 to 4, then 9 to 12, then 13 to 16. Their first proof should be a simple documented workflow they can run twice.
What worksCritique Before Copy, Claude for Thought Work, and Human Review Pass will make sense because creative output needs taste, audience fit, and revision.
RiskThey may overfocus on prompts that make prettier content and underfocus on source checks, version control, and repeatable publishing.
First methodStart with Claude Web for critique and ChatGPT Web for practice drafts. Move to Codex only when publishing the asset.
First pathStart at lessons 5 to 8, then 21 to 24, then 13 to 16. Their first proof should be one content asset with source notes, revision notes, and final QA.
What worksThis learner will like the Architect path because they already think in offers, customers, operations, and repeatable service delivery.
RiskThey may jump to building a personal AI operating system before learning verification, safe files, and clean handoffs.
First methodStart with Claude Web to map the process, then Codex or Claude Code when the process needs files, scripts, or a page.
First pathStart at lessons 1 to 4, then 17 to 20, then 29 to 32. Their first proof should be one business process mapped into intake, output, and QA.
What worksThe map gives them a lane. Instead of saying they can do anything, they can become the person who organizes messy work into verified outputs.
RiskThey may try to take every lesson in order and stall. They need a narrower work promise first.
First methodStart with Gemini Web or ChatGPT Web to compare lanes, then use Claude Web to turn one lane into an SOP.
First pathStart at lessons 6, 12, 15, 18, 25, and 26. Their first proof should be a cleaned tracker, source log, or SOP that shows organized thinking.
What worksThe blueprint is honest about the beginner stage. It does not ask them to pretend they are an AI expert on day one.
RiskThe full 32-lesson map may feel too big. They need a tiny first win before they can believe the rest is possible.
First methodStart with ChatGPT Web for practice or Gemini AI Mode for search-first learning. Avoid CLI tools until one small web-chat win is complete.
First pathStart only with lessons 1, 2, 3, and 5. Their first proof should be one simple prompt recipe and one small finished project, clearly labeled as practice.