Welcome back
Enter your name and PIN to load your progress across all your devices. New here? Just pick a name and PIN — we'll remember you.
Your progress syncs across all your devices. If you forget your PIN, start fresh with a new name.
Personal Roadmap · 2026 Edition

30-Day AI
Mastery Plan

From casual chat user to genuine AI power user — 2 to 3 hours a day

30Days
4Phases
~70hTotal
30+Deliverables
0 / 30 days done
Week One · Days 1–7 · 2 hrs/day
Master What You Already Have

You already use AI chat tools. This week you rebuild your entire relationship with them. Prompt engineering, personas, multimodal AI — by Sunday you'll get dramatically better results from tools you already pay for.

Expert prompting 25+ prompt library Custom AI personas Multimodal fluency Personal knowledge base
0 / 7
01
How LLMs Actually Work
Tokens · context windows · why AI hallucinates · model differences
+
Schedule — 2 hours
0:00–0:45Watch Karpathy's "Intro to LLMs" — the clearest explanation of how these models work that exists anywhere.
0:45–1:15Read Anthropic's model overview. Understand what a "context window" is — it explains 80% of AI behaviour. Do this before the comparison exercise so you know what to look for.
1:15–2:00Now run the same 5 prompts across Claude, ChatGPT, and Gemini. Write down every difference in tone, accuracy, and format. Use what you just read about context windows and model differences to guide your observations.
Resources
▶Intro to LLMs — Andrej Karpathy (YouTube)Video↗ 📄Claude Model Overview — Anthropic DocsDocs↗
Today's Deliverable
📝A one-page "model comparison" note: what each major AI is best and worst at, from your own live tests. Save this in your 30-Day folder.
02
Prompt Engineering Fundamentals
Zero-shot · few-shot · chain-of-thought · role prompting · output control
+
Schedule — 2 hours
0:00–0:45Read Anthropic's full prompt engineering guide. Take notes on each technique with a real example you wrote yourself.
0:45–1:30Take the 5 prompts from Day 1's comparison exercise and rewrite each one using a different technique from the guide you just read. Compare the new outputs side by side. Do this after Step 1 — you need the techniques fresh in your head before rewriting.
1:30–2:00Start your Prompt Library in Notion, Google Docs, or Obsidian. Add your 10 best prompts with categories.
Resources
📄Prompt Engineering Overview — Anthropic DocsDocs↗ 📚Learn Prompting — Free comprehensive course with exercisesGuide↗ 📚PromptingGuide.ai — Exhaustive technique referenceGuide↗
Additional Resources
📄Prompt Engineering Guide — OpenAI (if you want a second perspective)Docs↗
Today's Deliverable
📝Prompt Library v1 — 10 categorised, tested prompts you'll actually reuse weekly. This library grows every day this week.
03
System Prompts & Custom Personas
Custom instructions · persona design · XML structure · persistent context
+
Schedule — 2 hours
0:00–0:30Learn what system prompts are and why they change everything. Set up Claude's custom instructions with your background and preferences.
0:30–1:30Build 3 personas: "Writing Coach", "Business Advisor", "Code Reviewer". Test each with 5 real prompts and refine.
1:30–2:00Learn XML tagging for structured prompts — the most underused technique by non-developers. Try it on a complex task. This directly powers up the personas you just built — XML tags let you pass structured context into a system prompt cleanly.
Resources
📄System Prompts Deep Dive — Anthropic DocsDocs↗
Additional Resources
⭐Awesome ChatGPT Prompts — 200+ examples for inspirationGitHub↗
Today's Deliverable
📝3 saved personas set up in Claude (and ChatGPT if you use it) that you'll use from tomorrow onward. Not demos — actual tools.
04
AI for Your Actual Work
Document analysis · email drafting · research · summarisation · file uploads
+
Schedule — 2 hours
0:00–0:30Audit your last work week. List every task you did. Mark which were writing, research, analysis, or communication.
0:30–0:50Set up Claude Projects with your context — your role, company, working style. This becomes your base camp for the next 30 days. Do this before the real tasks below so the AI already knows who you are when you start.
0:50–2:00Now use AI on 3 real tasks from your audit list. Upload actual documents. Draft actual emails. Summarise a real report. No toy examples.
Additional Resources
🔬Google NotebookLM — You'll use this properly on Day 6Tool↗ 🔍Perplexity — AI research tool, covered properly on Day 11Tool↗
Today's Deliverable
📝A "Weekly AI Task Map" — a list of your recurring tasks with a note next to each one on how AI will now handle it, partially or fully.
05
Multimodal AI — Images, Voice & Video
Image generation · voice AI · video tools · visual prompting
+
Schedule — 2 hours
0:00–0:45Image generation: Try DALL-E 3 (inside ChatGPT) and Midjourney. Aim for around 10 images — not 20. The goal is learning what makes a strong vs weak visual prompt, not volume. Midjourney requires a Discord account — if you don't have one, DALL-E 3 alone is fine for today.
0:45–1:30Voice AI: ElevenLabs for text-to-speech, ChatGPT voice mode for two-way conversation. Transcribe a voice note with Whisper.
1:30–2:00Video: Explore Runway or Sora. Even generating a 5-second clip demonstrates the capability and burns it into memory.
Resources
🎨ChatGPT (with DALL-E 3) — Image generation, no extra signupTool↗ 🔊ElevenLabs — Best-in-class voice AI, free tier availableTool↗ 🎙️Whisper — OpenAI's speech-to-text (free, open source)Tool↗ 🎬Runway — AI video generationTool↗
Additional Resources
🎨Midjourney — Professional image generation (requires Discord)Tool↗
Today's Deliverable
📝One AI-generated image, one voice clip, and one video clip saved to your 30-Day folder. A tangible demo of your new range.
06
RAG & Personal Knowledge Bases
Retrieval-Augmented Generation · NotebookLM deep dive · second brain setup
+
Schedule — 2 hours
0:00–0:30Learn what RAG is: why "chatting with your documents" is different from a regular chatbot — and why it's far more accurate for your own content.
0:30–1:30Build your personal NotebookLM: upload your notes, PDFs, saved articles. Generate an "Audio Overview" of one topic you're studying.
1:30–2:00Expand your Claude Project with memory: paste your bio, goals, and working style. Test how much richer responses become with context.
Resources
🔬Google NotebookLM — Free, powerful RAG toolTool↗ ▶RAG Explained Simply — No coding requiredVideo↗
Today's Deliverable
📝A live NotebookLM loaded with your own documents that you can actually query. Ask it 5 real questions and screenshot the answers.
07
Week 1 Review & Prompt Library Polish
Consolidate · measure · expand library to 25 prompts · set Week 2 goals
+
Schedule — 2 hours
0:00–0:45Review all 6 deliverables from this week. Write a short "what surprised me most" reflection. Be honest.
0:45–1:30Expand your Prompt Library to 25 prompts. Add categories: Writing, Research, Analysis, Code, Creative, Work Tasks.
1:30–2:00Measure Week 1: how many hours did AI save you? Which tools do you now use daily? Write 3 goals for Week 2.
Week 1 Checkpoint
✓ Can you write a prompt that gets a near-perfect output on the first try? Do you have 25+ saved prompts? Can you explain what a context window is? Have you used AI on real work tasks this week? Yes to all → Week 1 complete. Move to Week 2.
Week Two · Days 8–14 · 2.5 hrs/day
Build Real Automations

Go beyond chat. This week you build things that run without you — live automations, API calls, AI-powered scripts. By Friday you'll have 3+ workflows saving real hours every week.

3+ live automations First API call AI-assisted coding Specialist tools integrated
0 / 7
08
AI Coding Without Being a Developer
Cursor · vibe coding · build a real mini-app · debugging with AI
+
Schedule — 2.5 hours
0:00–0:20Download and install Cursor from cursor.com. Create a free account and complete the initial setup wizard. Do this first — installation alone can take 10–15 minutes depending on your connection. Don't skip this or it eats into build time.
0:20–0:45Watch the Cursor intro walkthrough video. Key mindset shift: you're not learning to code, you're learning to direct an AI that codes for you.
0:45–1:45Build something in one session: a personal expense tracker, daily standup generator, or text formatter. Let AI write 100% of the code.
1:45–2:30Break it intentionally. Paste the error into Claude. Fix it. This debugging loop is the most valuable skill in AI-assisted development.
Resources
⌨️Cursor — AI-first code editor, free tierTool↗ ▶Build an App with Cursor — Beginner walkthroughVideo↗
Additional Resources
🚀Replit — Alternative if you want zero local setupTool↗
Today's Deliverable
📝One working mini-app built entirely with AI assistance. Take a screenshot. Write one sentence on what you'd build next.
09
No-Code Automation with Zapier
Triggers · actions · AI steps · connect your real tools
+
Schedule — 2.5 hours
0:00–0:45Learn the trigger → action → AI step model. Watch Zapier's intro to AI steps. Understand the pattern before you start building.
0:45–2:00Build 2 live Zaps: (1) New email → AI summarise → save to Notion. (2) Form response → AI classify by topic → add to spreadsheet with label.
2:00–2:30Document both Zaps: what problem each solves and how many minutes per week they save you. This is your automation portfolio starting.
Resources
⚡Zapier AI — Automation with built-in AI stepsTool↗
Additional Resources
⚙️Make.com — You'll use this properly on Day 12Tool↗
Today's Deliverable
📝2 live automations with documented time savings. If they don't run automatically by end of day, they don't count.
10
Your First API Call
What APIs are · Anthropic console · build a useful script with AI help
+
Schedule — 2.5 hours
0:00–0:30Plain-English: what is an API? Get your Anthropic API key from console.anthropic.com. This takes 5 minutes.
0:30–1:15Make your first API call from the Anthropic Workbench — no code required. Understand temperature, max_tokens, system prompts as parameters.
1:15–2:30With Cursor's help, build a Python script that calls Claude to do one useful thing — e.g., summarise any URL you paste in. This is a bigger jump than it looks. If you get stuck, tell Claude exactly what error you're seeing and ask it to fix the script. That back-and-forth IS the skill — don't expect it to work first try.
Resources
🔑Anthropic Console — Get API key + try Workbench freeTool↗ 📄Anthropic API Quickstart GuideDocs↗
Additional Resources
🔧Postman — Visual API testing tool if you prefer a GUITool↗
Today's Deliverable
📝A working script (written with AI help) that calls an AI API to automate one real task. You should be able to run it again tomorrow without help.
11
Specialist AI Tools
Gamma · Fireflies · Descript · Perplexity deep dive · Notion AI
+
Schedule — 2.5 hours
0:00–0:45Gamma: build a real presentation from a prompt on a topic you actually care about. Notice how it structures content and where to edit vs accept.
0:45–1:30Fireflies: set up AI meeting recording and summaries. Run a test meeting. Check the transcript quality and action items generated.
1:30–2:30Perplexity: use it for 3 real research tasks you'd normally Google. Compare depth, citation quality, and time spent versus standard search.
Resources
📊Gamma — AI presentations in minutesTool↗ 🎙️Fireflies — AI meeting notes and action itemsTool↗ 🔍Perplexity — AI research with live citationsTool↗
Additional Resources
🎬Descript — Edit video by editing text (explore if relevant to your work)Tool↗
Today's Deliverable
📝One AI-generated presentation on a real topic. A note on which specialist tools you'll keep using weekly and which you're dropping.
12
Advanced Make.com Automation
Multi-step scenarios · webhooks · error handling · complex flows
+
Schedule — 2.5 hours
0:00–0:45Make.com intro: understand scenarios, modules, and data mapping. More powerful than Zapier for complex multi-step flows.
0:45–2:00Build a 3+ step scenario: new RSS article → AI summarise → add to Notion database → send Slack notification.
2:00–2:30Learn webhooks: how to trigger automations from external events. This concept is essential for the agent work in Week 3. On Day 16 you'll build an n8n agent that uses exactly this pattern — a webhook triggers the agent to start a multi-step task. The mental model you build today carries directly there.
Resources
⚙️Make.com — Free tier, visual automation builderTool↗ ▶Make.com Official YouTube — Full tutorial libraryVideo↗
Today's Deliverable
📝One multi-step Make scenario with 3+ modules, running on a schedule or webhook trigger. Document it in your portfolio.
13
AI Stack Audit
Map every tool · cut what's not working · identify gaps · optimise
+
Schedule — 2 hours
0:00–0:45Identify 3 remaining workflow gaps in your life and work. For each, research whether an existing AI tool already solves it well. Note the best candidates. Do this before drawing the diagram — you can't accurately map what you have until you've reflected on what's missing.
0:45–1:30Now draw your current AI stack — including the new tools you just found. A simple diagram: every tool, what it does, how often you use it, what it replaced. Be brutal about what's not sticking.
1:30–2:00Calculate your total weekly time savings so far. Update the Prompt Library. Write a one-paragraph preview of what you want to build in Week 4.
Today's Deliverable
📝Your "Personal AI Stack" diagram — a clean visual of every tool you now use regularly and what manual task each one replaced.
14
Week 2 Review & Automation Showcase
Fix what's broken · document savings · prep for agents
+
Schedule — 2 hours
0:00–1:00Run through every automation you've built. Fix anything that's broken or unreliable. Improve ones that work but could be smarter.
1:00–1:30Complete your automation portfolio: name, purpose, trigger, estimated hours saved per week. This becomes proof of your capability.
1:30–2:00Week 2 checkpoint self-assessment. Decide which agent you'll build first in Week 3.
Week 2 Checkpoint
✓ Have you built 3+ automations that are still running? Called an AI API successfully? Built something with Cursor? Are you measurably saving hours per week? Yes to all → you're ready for agents. Move to Week 3.
Week Three · Days 15–21 · 2.5–3 hrs/day
Agents & OpenClaw

This is the week everything clicks. You'll understand agents deeply, build one with n8n, and deploy OpenClaw — your always-on AI that lives on your machine, connects to your messaging apps, and actually does things autonomously. Security-first throughout.

Agent architecture n8n agent live OpenClaw deployed Persistent AI memory Security baseline
0 / 7
15
What Are AI Agents?
Agent anatomy · tools · memory · planning loops · real-world examples
+
Schedule — 2.5 hours
0:00–0:45Explore live examples first: Claude computer use demo, a public Auto-GPT instance. Try to break them. Notice what fails and ask yourself why. Seeing a real agent before reading the theory makes the framework click much faster — you'll recognise the concepts as you read them.
0:45–1:30Now read Anthropic's "Building Effective Agents." You'll understand it much better having just seen a real agent in action. Focus on the sections on tools, memory, and failure modes.
1:30–2:30Design your ideal agent on paper: the problem it solves, tools it needs, possible failure modes. This spec guides the next two days.
Resources
📄Building Effective Agents — Anthropic Research (essential read)Read↗ 🖥️Claude Computer Use — Live agent demoDocs↗
Today's Deliverable
📝Written "Agent Spec" — one page: the problem, tools needed, trigger, expected output, and top 3 failure modes you'd need to guard against.
16
Build Your First Agent with n8n
Multi-step autonomous workflows · tool use · loops · real deployment
+
Schedule — 3 hours
0:00–0:30Sign up and set up n8n cloud free tier. Complete their intro workflow tutorial to understand the interface. Sign up the night before if you can — email verification and onboarding can eat into this 30 minutes. If setup bleeds over, trim from the build time, not the testing time.
0:30–2:30Build the agent from your Day 15 spec. Good starter: a Research Agent that takes a topic, searches the web, summarises findings, and emails you a briefing. The webhook concept from Day 12 is the trigger pattern you'll use here — refer back to those notes.
2:30–3:00Test it 5 times with different inputs. Document every failure. Fix at least 2 before moving on.
Resources
🤖n8n — Open-source agent and automation builderTool↗ ▶Build an AI Agent with n8n — Beginner walkthroughVideo↗
Today's Deliverable
📝One live AI agent completing a multi-step task without you manually triggering each step. Screenshot it running successfully.
17
OpenClaw — Preparation & Security
Architecture · prerequisites · security policy · what NOT to give it access to yet
+
🦞 About OpenClaw
OpenClaw is a self-hosted AI agent that runs on your machine 24/7, connecting to your messaging apps (WhatsApp, Telegram, Discord), files, browser, and shell commands. It has persistent memory, cron jobs, and can write its own skills. The excitement is legitimate — and so is the need for care.
⚠️ Security: OpenClaw can run shell commands and access your files. Cisco researchers found third-party skills can exfiltrate data silently. Always start with least privilege — read-only access first, expand only what you actively need.
Schedule — 2.5 hours
0:00–0:45Read the OpenClaw docs: architecture overview and Getting Started. Read the Cisco security analysis. Do not install yet.
0:45–1:30Plan your setup: which messaging app (Telegram recommended), what will you allow access to, what is strictly off limits.
1:30–2:30Set up prerequisites: Node.js, a secondary Telegram account (not your main number), and your Claude API key ready.
Resources
📄OpenClaw Personal Assistant Setup GuideDocs↗ ⭐OpenClaw GitHub — Source code and full docsGitHub↗ 🔒Cisco: OpenClaw Security Analysis — Read before installingRead↗
Today's Deliverable
📝A written security policy for your OpenClaw instance: what it can access, what it cannot, which community skills are allowed. One page. This is your guardrail document.
18
OpenClaw — Install & First Conversation
Run onboard wizard · write SOUL.md & USER.md · first 10 tests
+
Schedule — 3 hours
0:00–1:00Run openclaw onboard. Follow every step. Connect Telegram as your channel. Connect to Claude via API key.
1:00–1:45Write SOUL.md (your assistant's persona and tone) and USER.md (who you are, your goals, working style, context). This is what gives it real memory of you.
1:45–2:00Run openclaw doctor and fix any warnings before testing anything. Do this before the testing step — running 10 requests on a misconfigured instance wastes your time and can produce misleading results.
2:00–2:45Now test with 10 different requests: summarise something, set a reminder, look something up, ask for help with a task. Note what works and what fails.
2:45–3:00Fix the top 2 things that failed. Update USER.md with anything it misunderstood.
Resources
📄OpenClaw Full DocumentationDocs↗ 💬OpenClaw Discord — Community help when you get stuckCommunity↗
Today's Deliverable
📝OpenClaw running and connected to Telegram, with a written SOUL.md and USER.md. Send yourself a message from it as proof it works.
19
OpenClaw — Skills & Integrations
Add safe tools · notes · reminders · configure daily heartbeat
+
Schedule — 2.5 hours
0:00–0:45Add your first safe integrations: Notion or Apple Notes (read-write), Apple Reminders or Things (read-write), Calendar (read-only to start).
0:45–1:30Configure heartbeats: write your HEARTBEAT.md with a daily briefing template — what you want it to check and report each morning.
1:30–2:30Browse the community skills registry. For any skill you want to add, read its code or description before installing. Never install blindly. Refer back to the security policy you wrote on Day 17 before installing anything. If a skill asks for permissions not on your allowed list, skip it — no matter how useful it looks.
Additional Resources
📖Real OpenClaw Setup Story — Lessons learned (Medium)Read↗ 📖Using OpenClaw as a Personal Assistant — Towards Data ScienceRead↗
Today's Deliverable
📝OpenClaw with 2+ integrations and a working daily heartbeat that sends you a morning briefing. Screenshot it working.
20
OpenClaw — Real Workflows & Debugging
Build one workflow you'd actually miss · fix memory issues · iterate
+
Schedule — 2.5 hours
0:00–1:00Build one real workflow: e.g., forward newsletters to a dedicated inbox → OpenClaw summarises daily and sends you the highlights. Or: daily goal log via voice message.
1:00–1:45Debugging: what has it gotten wrong this week? Fix memory issues, rewrite parts of USER.md, clarify which skills to use in which situations.
1:45–2:30Learn the key principle: "reach for a script before you reach for a prompt." Build one deterministic script for a task that needs to happen consistently and exactly.
Today's Deliverable
📝One real workflow running through OpenClaw that you'd notice immediately if it stopped. Document it: what it does, when it triggers, what it sends you.
21
Week 3 Review & Security Audit
Review all agents · run openclaw doctor · trim permissions · plan Week 4
+
Schedule — 2.5 hours
0:00–0:45Security audit: run openclaw doctor, review all current permissions, remove any access you haven't actively used.
0:45–1:30Document your agent portfolio: n8n agent, OpenClaw setup, key workflows. What does each do? What would you replace it with if it stopped working?
1:30–2:30Week 3 checkpoint + decide your Week 4 capstone project. Pick something ambitious but achievable in 9 days. Write a 3-sentence description of it.
Week 3 Checkpoint
✓ Is OpenClaw running and giving you daily value? Have you built an n8n agent? Can you explain the difference between an automation and an agent? Have you done a security audit? Yes to all → you're an AI power user. Week 4 turns you into a builder.
Week Four · Days 22–30 · 2.5–3 hrs/day
Build Something Real & Ship It

The capstone week. Study how people are building AI businesses, pick a project, build it using everything from the past three weeks, and get it in front of real people by Day 28. The final two days are for iteration and planning what comes next.

AI business models Live project shipped Real user feedback 90-day roadmap
0 / 9
22
AI Business Models & Opportunities
How people make money · micro-SaaS · services · prompt products · real examples
+
Schedule — 2.5 hours
0:00–1:00Study 5 real AI businesses on Indie Hackers. For each: what it does, how it's built, what it charges, who pays. Look for patterns across all 5.
1:00–1:45Map your own opportunity: which problems in your industry or daily life could an AI product solve? Generate 10 ideas, ruthlessly narrow to 3. Use the patterns you spotted in Step 1 as your filter — if none of the 5 businesses you studied would do what you're thinking, that's either a gap or a warning sign. Figure out which.
1:45–2:30For your top idea: use Claude to research competitors, estimate market size, and draft a one-page business case.
Resources
💡Indie Hackers — Real AI founder revenue storiesSite↗ 🗺️There's An AI For That — Spot gaps in the marketSite↗ 🚀Product Hunt AI — See what's launching right nowSite↗
Today's Deliverable
📝One-page business case for your chosen idea: the problem, target customer, AI-powered solution, revenue model, and top competitor.
23
Project Scoping & Architecture
Finalise what you're building · design the stack · create a 5-day build plan
+
Schedule — 2.5 hours
0:00–0:45Finalise your project — it should flow directly from the business case you wrote on Day 22, not be a fresh idea. Good options: an AI chatbot for a niche, an automation service for a business type, a prompt pack to sell on Gumroad, or a personal tool that solves your own real problem. If you're tempted to pick something completely different from Day 22's research, ask yourself why — pivot if there's a real reason, but don't restart just because it feels easier.
0:45–1:30Design the stack: which tools from Weeks 1–3 will you combine? Where does the AI live? What does a first-time user actually do?
1:30–2:30Write your 5-day build plan: Days 24–28, each with one specific milestone. Work backwards from Day 28 — that's launch day. If a milestone isn't achievable in the time, cut scope now, not on Day 27.
Today's Deliverable
📝Project brief: what you're building, the full stack, and a milestone for each of the next 5 days. Day 28 = live URL. Commit to this.
24
Build Day 1 — Core Functionality
Get the core working · AI does the heavy lifting · don't polish yet
+
Schedule — 3 hours
0:00–2:30Pure build. Core feature only. Use Cursor for code, Claude for copy and logic, your API knowledge for AI features. The rule: make it work before you make it beautiful.
2:30–3:00Test it yourself 10 times with real inputs. Write down every bug and every "this could be better" note for tomorrow.
Resources
▲Vercel — Deploy web apps free, instantlyTool↗ 🚀Replit — Build and host, zero local setupTool↗
Additional Resources
💰Gumroad — For selling digital products (relevant from Day 28 onward)Tool↗
Today's Deliverable
📝Core functionality working on your local machine or in Replit. A written bug list for Days 25–26.
25
Build Day 2 — Fix & Deploy
Fix top 5 bugs · add one more feature · get it live with a URL
+
Schedule — 3 hours
0:00–1:30Fix the top 5 bugs from yesterday. Paste error messages verbatim into Claude — never paraphrase errors when debugging.
1:30–2:30Add one more feature or improve the core UX. Ask: "what would make a stranger understand what this does in 10 seconds?"
2:30–3:00Deploy to Vercel or Replit. Even if rough — get it online so you have a URL you can share tomorrow.
Resources
▲Vercel — Deploy web apps free, instantlyTool↗ 🚀Replit — Alternative host, zero local setup neededTool↗
Today's Deliverable
📝A live URL. Something you could show someone right now and have them understand what it does within 10 seconds.
26
Build Day 3 — Real User Testing
Share with 3 honest people · watch them use it · fix the embarrassing parts
+
Schedule — 3 hours
0:00–0:30Share the URL with 3 people who'll be honest. Don't explain it first — see if they can figure out what it does on their own.
0:30–2:00Watch how they use it (ask them to share their screen or describe what they're doing). Fix the 3 most embarrassing problems immediately.
2:00–3:00Write your copy: a one-sentence description, a headline, and 3 bullet points on what it does. Use AI to help but make every word honest.
Resources
📊Gamma — Quickly draft a landing page or one-pager with AITool↗
Today's Deliverable
📝Feedback notes from 3 real people and a list of exactly 3 things you fixed as a direct result of what they said.
27
Build Day 4 — Polish & Launch Prep
Nail the first impression · write landing page · prepare 3 launch posts
+
Schedule — 3 hours
0:00–1:30Polish the core experience. Ask yourself: if someone used this once and never came back, what one thing would you fix? Fix that thing.
1:30–2:30Write a simple landing page or README: what it does, who it's for, how to use it. Use Gamma or plain HTML — keep it to one screen.
2:30–3:00Write 3 launch posts: one for LinkedIn, one for X/Twitter, one for a relevant Reddit or community. Different tone, same truth.
Resources
📊Gamma — Build your landing page from a prompt in minutesTool↗ ▲Vercel — Deploy and host your landing page freeTool↗
Today's Deliverable
📝A live, polished project with a landing page. 3 launch posts drafted, proofread, and ready to go tomorrow morning.
28
Launch Day 🚀
Ship · post · share · collect feedback · don't hide
+
Schedule — 2.5 hours
0:00–0:30Final check: does it work end-to-end right now? Does the landing page explain it clearly? Is the URL clean and shareable? If something is fundamentally broken, don't launch today — take the extra time to fix it. A delayed launch is fine. A broken launch that turns off your first users is hard to recover from.
0:30–1:00Post it everywhere. Use your best launch post. Share in 2–3 relevant communities. Send the link directly to 5 people with a personal note.
1:00–2:30Engage with every response, comment, and reaction today. Document everything — positive and negative. This feedback is your most valuable asset.
Today's Deliverable
🚀Something live on the internet that you built with AI, shared publicly, and received real feedback on. This is the deliverable that proves everything else.
29
Post-Launch Iteration
Process feedback · fix top issues · ship v1.1 · plan Version 2
+
Schedule — 2.5 hours
0:00–0:45Process all launch feedback. Group it into: bugs, feature requests, confusion points, and positive signals. Find the strongest pattern.
0:45–1:45Fix the top 3 issues. Ship an update. Post "v1.1" back to wherever you launched. Showing you respond to feedback matters as much as the fix itself.
1:45–2:30Write your Version 2 spec: what's the most important thing to build based on what users actually said? One page. Prioritised.
Today's Deliverable
📝Version 1.1 live with at least 3 user-driven improvements. A written V2 spec based entirely on what real people told you.
30
Graduation — Review & 90-Day Plan
Full programme review · measure your growth · staying-current system · what's next
+
Schedule — 3 hours
0:00–1:00Full review: re-read every weekly checkpoint and all 30 deliverables. Write an honest "Day 1 vs Day 30" comparison — what can you do now that you couldn't before?
1:00–1:45Set up your staying-current system: subscribe to 3 newsletters, follow 5 AI builders, join 1 community. AI moves fast — your information edge matters.
1:45–3:00Write your 90-day plan: 3 concrete projects or skills, monthly milestones, one stretch goal. Use Claude to pressure-test it — ask it to find the weak spots in your plan.
Stay Current
📰Ben's Bites — Best daily AI newsletterNewsletter↗ 📰The Rundown AI — Daily digest, clear and conciseNewsletter↗ 🎙️Latent Space Podcast — For serious AI buildersPodcast↗ 👥Indie Hackers — Build in public communityCommunity↗
Final Checkpoint
🎓 You have: a 50+ prompt library · 6+ automations saving hours weekly · an n8n agent running · OpenClaw deployed and useful · something live on the internet with real user feedback · a 90-day plan. You started as a chat user. You are now an AI builder. That's the whole point.
Track Your Progress

Six metrics that tell you whether you're actually learning or just going through the motions.

📚
Prompt Library
Saved, categorised, reusable prompts. Quality matters more than count.
25+ by Day 7 · 50+ by Day 30
⚡
Live Automations
Workflows actively running without you. Only count ones still working.
3+ by Day 14 · 6+ by Day 30
🕐
Hours Saved / Week
Real time no longer spent on tasks now handled by AI. Track it honestly.
3+ hrs/wk by Day 14 · 7+ by Day 30
🤖
Agents Deployed
Multi-step autonomous workflows including your OpenClaw setup.
n8n agent by Day 16 · OpenClaw by Day 18
🛠️
Tools in Daily Stack
AI tools integrated into actual daily work — not just signed up for.
6 by Day 14 · 10+ by Day 30
🚀
Things Shipped
Anything built with AI and shared publicly. One real thing beats ten demos.
1 live project by Day 28 · V2 by Day 30
Your 30-Day AI Stack

Every tool you'll encounter across the programme, with honest pricing labels.

Claude
Best for long context, deep reasoning, writing, and Projects memory.
Freemium
ChatGPT
Image generation, voice mode, plugins, and broad general use.
Freemium
Perplexity
AI research with live web citations. Replaces search for deep queries.
Freemium
NotebookLM
Chat with your own documents. Best free RAG tool available.
Free
Cursor
AI-first code editor. Build real apps without knowing how to code.
Freemium
Zapier
No-code automation with 5,000+ app integrations and AI steps.
Freemium
Make.com
Visual automation builder. More powerful than Zapier for complex flows.
Freemium
n8n
Open-source agent and automation builder. Self-host for free.
Free (self-host)
OpenClaw 🦞
Self-hosted 24/7 AI personal assistant on your own machine.
Free + API costs
ElevenLabs
Best-in-class AI voice generation. Realistic text-to-speech.
Freemium
Gamma
AI-generated presentations and documents in minutes.
Freemium
Vercel
Deploy web apps instantly, free. Your home for everything you ship.
Free tier
Bonus Section · Fund Your AI Usage
Make AI Pay for Itself

Heavy AI usage with API access costs $20–80/month for a power user. You don't need a business — you need one small income stream. Here's exactly how to build it using the skills from this plan.

What you're actually funding

The numbers are smaller than you think.

Claude API (heavy use)
~$30
per month for serious daily usage via API
OpenClaw 24/7
~$20–40
per month running continuously on Claude Sonnet
Full power stack
~$80
per month worst case: API + tools + subscriptions
What you need to earn
$100
per month covers everything with buffer to spare
Four income streams — from easiest to most scalable

Pick one. Do it properly. One stream covering $100/month is all you need.

⚡
Automation Setup for Local Businesses
Small businesses — restaurants, salons, real estate agents, retailers — need automations but have no idea how to build them. A Zapier workflow handling inquiry emails, a Make scenario scheduling social posts, a tool summarising customer reviews. You charge a one-time setup fee. You have these skills by Day 14.

Approach 3 local businesses. Show them a specific problem you can solve in a demo. Charge for the setup, offer a small monthly retainer for maintenance.
Fastest to money
$150–500
per client, one-time
1 client = 6 months covered
📦
Prompt Packs on Gumroad
A curated set of 25–40 prompts for a specific profession sells for $9–29. Lawyers, teachers, HR managers, marketers, real estate agents — every profession has repetitive writing tasks that great prompts dramatically speed up. You build a prompt library in Week 1. Packaging it as a product takes one extra afternoon.

Pick one profession you understand. Write 30 prompts. Test them. Add a one-page PDF guide. List on Gumroad. Share in 2–3 relevant communities or subreddits.
Passive once live
$9–29
per sale
5–10 sales/month = covered
🎯
AI Setup Consulting Sessions
By Week 3 you know more about practical AI than 95% of people. A 90-minute "AI setup session" — helping someone configure Claude Projects, set up their first automation, build a prompt library — is genuinely worth paying for. People spend hours confused by things you can now solve in minutes.

Create a simple Calendly booking page. Offer a 90-minute session. Write one LinkedIn or X post explaining what you do. You only need 1–2 bookings a month.
Skills-based
$75–150
per session
1 session/month = covered
🚀
Your Week 4 Project With a Payment Link
The most aligned option: your Day 28 capstone project already has to be something real. Aim it at a paying use case from the start. A niche AI tool with a $5–15 one-time purchase, a template pack, an automation service with a monthly fee. The project you'd build anyway — just with a Gumroad or Stripe link attached.

This is the only option that directly turns your learning into income. Even 10 sales at $10 covers two months of API costs and proves the model works.
Most scalable
$5–15
per sale or /month
compounds over time
When to do what — layered over your 30 days
Week 1–2
Focus on learning only. No monetisation pressure. Build skills genuinely — you can't sell what you don't understand yet.
Day 14
Identify one local business in your network who'd pay for an automation you can now build in a day. Have an informal conversation — don't pitch yet, just ask about their pain points.
Day 20
Package your prompt library as a product on Gumroad. List it. Share it once. Even $0 earned tells you something about positioning — iterate the description.
Day 23
Scope your capstone with revenue in mind. Whatever you build in Week 4 should have a payment link from day one. Not as an afterthought — as part of the brief.
Day 28+
Your first income experiment is live. It might make $10 or $100. Either way, AI costs are now a business expense — not personal spending. That mental shift is worth more than the money.
Platforms to sell on
💰Gumroad — Sell digital products, zero upfront costSell↗ 📅Calendly — Paid booking page for consulting sessionsSell↗ 💳Stripe — Add a payment link to any projectPayments↗ 🎯Fiverr — List AI automation services, inbound leadsMarketplace↗ 👥Indie Hackers — Community of people building exactly thisCommunity↗
The 30-Day AI Mastery Plan
From chat user to AI builder — 2 hours a day, 30 days, one project shipped.

Share this with anyone who feels left behind by AI.