Most people are still using AI like a search box. They open a chat window, ask a question, copy the answer, and move on. AI-native people use it like an operating system: to research faster, write sharper, analyze data, make visuals and video, prototype ideas, automate the boring parts, and make better calls. The gap between the two is not access to expensive tools. It is practice, workflows, judgment, and proof.
This is a 30-day plan to close that gap. The promise is simple: in a month, working a little each day, you will learn the major AI tools, complete free or freemium courses, build practical workflows, and publish a portfolio that shows you can use AI in real work.
One honest caveat up front. This is not an official certification from OpenAI, Google, Anthropic, Stanford, or MIT. It is a practical proof-of-work challenge. Some of the courses below hand out completion certificates or badges, and a few of those cost money. The goal here is not a line on a résumé you paid for. It is to become genuinely AI-native, measured by visible skills and the artifacts you ship.
And the price, since that is the whole point. A typical AI bootcamp or certificate program charges $500 to $3,000 for roughly what this plan covers: foundations, productivity workflows, and building a working agent. This plan costs $0 and about 35 to 40 hours for the daily plan spread over a month — plus your chosen course track, which adds anywhere from a couple of hours to twenty-plus depending on the level you pick — and you can still collect real certificates along the way — Anthropic Academy, the Hugging Face Agents Course, OpenAI Academy, Google Skills, and IBM SkillsBuild all award them free. Certificates are what paid programs monetize. The portfolio you ship is what actually convinces anyone.
How this challenge works
Most "learn AI in 30 days" plans are a tour of tools you forget in a week. They hand you a list of apps, drill prompting, and skip the part that actually separates fluent people from dabblers: judgment. Anthropic's AI Fluency framework is a useful tell here. It breaks the skill into four competencies, the "four Ds": Delegation (knowing what to hand off), Description (prompting), Discernment (checking the output), and Diligence (using it responsibly). Prompting is one of four. Most plans spend all month on it.
So this plan runs on four phases instead of a tool tour — Judgment, Output, Creation, Systems — and three rules run through every day:
- Proof over completion. You finish by shipping ten artifacts, not by ticking boxes. A public portfolio beats a paid certificate you can't demonstrate.
- Verify, don't trust. Free tiers, credits, and certificate rules change constantly, and models still make things up. People also badly overestimate what AI does for them: a study of 25,000 Danish workers found AI users' measured time savings averaged about 3%, and in METR's 2025 randomized trial, experienced developers believed AI made them ~20% faster while actually finishing 19% slower. So every day includes checking the output and confirming the current terms yourself. That is why the labels below include a
VERIFYtag and why we keep a live AI tools directory and rate the hype so you are not taking a vendor's word for it. - Real work, not toy exercises. Each day maps to something you actually do at your job. We publish step-by-step AI workflows for exactly this reason, and this plan borrows the same "does it help in your next work session" test.
What does "AI-native" actually mean at the end of this? Not knowing a handful of ChatGPT prompts. It means you can:
- choose the right tool, model, and tier for the job instead of forcing everything through one chatbot;
- turn vague, one-off work into repeatable workflows, and keep a prompt library that compounds;
- verify outputs instead of trusting them blindly, and know what you are and are not allowed to paste into a tool;
- produce portfolio artifacts that prove real capability.
The outcome fits in one sentence: I can use AI to produce better work, faster, with judgment.
The daily rhythm: 15 minutes learning, 45 minutes building, 15 minutes reflecting or verifying. The building is the point — if you only have time for one part on a given day, keep the 45 minutes of building. Every day's table below pairs the learn slot with its own direct link, so you never have to hunt through a course catalog. And the build days that need a running start carry a ▶ worked example — a full recipe with copy-paste prompts on our workflows hub — so you're never staring at a blank screen wondering what "build an agent" actually means. Two honest notes on the math: your structured course track (pick one in the next section) is a multi-hour commitment you chip away at across the whole month, not a 15-minute-a-day thing; and the biggest Phase 4 build days — the prototype and agent builds on Days 25–26 — can run 90 to 120 minutes, so plan for that or split them across two sittings.
Free first, upgrade only when it earns it. Every day can be completed with a free course, a free plan, a freemium tool, or a no-cost alternative. One 2026 reality to plan around: a lot of "free" tiers have quietly become one-time credit buckets — enough to try, not enough to run. Treat every free tier as a trial and confirm the current limits before you build anything on it. Four labels keep this honest throughout:
| Label | Meaning |
|---|---|
| FREE | Can be used or completed without payment |
| FREEMIUM | Has a free tier, but limits apply |
| OPTIONAL PAID | Useful, but not required |
| VERIFY | Pricing, access, credits, and certificate rules may change |
Start here: pick your level
The 30 days are the same spine for everyone, but you should run one serious course track alongside them for depth. Pick your level, start its first course on Day 1, and chip away all month. The [full directory at the end of this article](#the-course-and-tool-directory) has every course with its own direct link, time commitment, and what "free" actually includes — or browse our curated free courses for more.
| You are | Run the 30 days like this | Your course track |
|---|---|---|
| Non-technical professional — executive, consultant, marketer, operator | Follow every day as written; take the no-code lane in Phase 4 | [Level 1: AI literacy](#level-1-ai-literacy-for-everyone) — start with OpenAI's AI Foundations, add Ng's AI for Everyone |
| Builder / operator — PM, analyst, creator, founder, semi-technical | Follow every day; choose your Phase 4 lane by comfort, not job title | [Level 2: builder & workflow operator](#level-2-ai-builder-and-workflow-operator) — Anthropic Academy and the agent courses |
| Engineer / technical — developer, data, technical founder | Compress Phases 1–2 into the exercises, spend the saved time in Phase 4's code lane | [Level 3: technical foundation](#level-3-technical-ai-foundation) — Ng's ML Specialization, then pick a lane |
Day 0: measure your baseline
"Proficient" only means something if you can show movement. Before Day 1, score yourself 1 to 5 on each statement below (1 = not at all, 5 = confidently, with an example you could show someone). The four map to Anthropic's four Ds. Retake it on Day 30 and publish both scores next to your portfolio — a documented before-and-after is more convincing than any badge, including the one this challenge hands out.
- Delegation. I can look at my week and tell which tasks to hand to AI and which to keep human. I can pick the right model and tier for a task instead of forcing everything through one chatbot.
- Description. I can write a prompt with role, context, task, constraints, examples, and format, and iterate on a weak output methodically instead of re-rolling and hoping. I keep a reusable prompt library.
- Discernment. I define what "good" looks like before I prompt and check outputs against sources. I have caught and documented at least one confident-but-wrong AI answer.
- Diligence. I know exactly what I can and cannot paste into each tool I use, keep a human approval step on anything high-impact, and can explain prompt injection and over-permissioning well enough to check my own agent for both.
Rough read: mostly 1s and 2s is a search-box user, mostly 3s is AI-assisted, mostly 4s and 5s is AI-native. The four phases below are how you move up.
Phase 1 — Judgment (Week 1: Days 1–7)
By Day 7 you can: write structured prompts and keep them in a library that compounds; pick the right model and tier for a task and know what it costs; fact-check AI output against a rubric you built; pull cited answers out of your own documents; and say exactly what you can and cannot paste into a tool. Seven artifacts prove it, starting with your Opportunity Map.
| Day | Focus | Today's lesson | Build (45 min) | Artifact |
|---|---|---|---|---|
| 1 | AI baseline & opportunity map | Start AI Foundations (OpenAI Academy, ~70 min — finish over Days 1–2). Optional deep-dive this week: Ng's AI for Everyone (~7 hrs) | List 20 weekly tasks AI could improve, score them by frequency, time, and stakes, then pick 3 to attack this month | AI Opportunity Map |
| 2 | Prompting basics | AI Prompting for Everyone, the opening lessons | Rewrite 10 weak prompts using role, context, task, constraints, examples, format · ▶ worked example | Prompt Library v1 |
| 3 | Model & tier selection | AI Fluency, the Delegation section, alongside Tokens, explained | Run one task across models and tiers; score quality vs speed vs cost | Model & Cost Scorecard |
| 4 | Verification & the discernment loop | AI Fluency, the Discernment section | Define what "good" looks like before you prompt, fact-check against sources, and turn it into a reusable check rubric | Verification Rubric |
| 5 | Ground it on your own docs | NotebookLM's getting-started guide (genuinely short) | Clean and label a set of your own documents, load them, force cited answers, and compare against an ungrounded reply | Grounded Research Brief |
| 6 | Responsible AI & data boundaries | AI Foundations, the responsible-use lesson, plus your org's AI policy | Write what you can and can't paste (consumer vs enterprise tier, PII/PHI/source/roadmap, opt-out and retention), plus your rules for bias, copyright, and disclosing synthetic media | Responsible-AI & Data-Boundary Playbook |
| 7 | 🛠 Apply day — Phase 1 capstone | No new learning | Assemble your personal guide to safe, effective AI use and consolidate the prompt library you started on Day 2 | Personal AI Playbook v1 |
Day 6 is not optional throat-clearing. More than a quarter of what people paste into consumer AI tools now contains sensitive data, and the consumer tier is exactly where that leaks. Knowing the boundary is a core AI-native skill, not a compliance afterthought.
Phase 2 — Output (Week 2: Days 8–14)
By Day 14 you can: run your email, writing, meeting notes, document analysis, spreadsheets, and slide decks through AI workflows you built — with verification baked in — and you've packaged the whole thing into a daily system you'll actually keep using (the AI Chief of Staff Pack).
| Day | Focus | Today's lesson | Build (45 min) | Artifact |
|---|---|---|---|---|
| 8 | Email & communication | Applied AI Foundations (OpenAI Academy), the workflow lessons | Build prompts for replies, follow-ups, summaries, tone shifts, hard conversations, and save them to your library · ▶ worked example | Email Prompt Bank |
| 9 | Writing with AI | Applied AI Foundations, the drafting and revision material | Draft and revise a memo, article, or newsletter; spot-check facts with your Day 4 rubric | Polished Written Asset |
| 10 | Meeting workflows | Google's Generative AI Fundamentals path, the first module | Turn messy notes into agenda, decisions, risks, next steps · ▶ worked example | Meeting Operating System |
| 11 | Document analysis | Claude 101, the working-with-documents section | Analyze a report, PDF, transcript, or policy doc, grounded and cited | Executive Summary |
| 12 | Spreadsheet thinking | Gemini's AI formulas in Google Sheets | Have AI build formulas, find patterns, suggest charts, and verify every number | Mini Data Analysis |
| 13 | Presentations | Gamma's AI-presentation quickstart or Canva's Magic Presentations guide | Turn your research brief into a 7-slide deck with speaker notes | AI-Assisted Slide Deck |
| 14 | 🛠 Apply day — Phase 2 capstone | No new learning | Package your prompts and templates into a daily system and finalize your prompt library | AI Chief of Staff Pack |
Pick a role track after Day 14
The first two phases are the same for everyone. Once you have the foundations and a productivity system, aim the second half at your actual job. (This role track is separate from your Level 1/2/3 course track — that one you picked at the start.) Choose the hub that matches your role and follow it for workflows and tool breakdowns built for that job:
- Consulting & Enterprise — client intake, research memos, strategy decks; final project: a client advisory pack.
- Finance — close checklists, variance memos, board decks; keep uncleaned financials out of uncleared tools.
- Healthcare — literature triage, de-identified drafts; PHI never touches a tool without approval and a clinician in the loop.
- Legal — clause comparison, first-pass redlines; mandatory human verification and privilege checks.
- Human Resources — JD drafting, interview kits, policy Q&A; a human stays in every decision.
- Developer — codebase onboarding, tests, refactors; final project: an agentic coding demo with a written diff review.
- Founder & Operator — market research, landing pages, outreach; final project: an AI-assisted go-to-market system.
- Sales & Marketing — prospect research, personalized outreach, a repurposing engine; final project: an AI revenue assistant.
Not one of these? Start with Founder & Operator or Developer, the two most general tracks, then borrow from the rest. Students and career switchers should run the Developer or Founder track and make the public portfolio their final project.
Phase 3 — Creation (Week 3: Days 15–21)
By Day 21 you can: produce real design, video, and campaign assets with AI; repurpose one idea across every channel and stand up a reusable sales/support kit; run a multi-step deep-research report and catch its errors; and you have a custom assistant, grounded on your own documents, that you use every day — plus a full mini campaign shipped as proof.
| Day | Focus | Today's lesson | Build (45 min) | Artifact |
|---|---|---|---|---|
| 15 | AI visuals & design | Canva Design School's AI-presentations lesson | Create a graphic, carousel, and one-pager, plus five visual concepts for one idea | AI Design Kit |
| 16 | AI video & short-form | HeyGen's first-video guide or Descript's getting-started | Make one 30–60s explainer and repurpose it into a script, captions, and clips | Short-Form Video Pack |
| 17 | 🛠 Content repurposing | No new course — reread your Day 4 verification rubric before you repurpose anything | Turn one research brief into a newsletter, LinkedIn post, carousel, script, and X thread | Content Repurposing System |
| 18 | Sales & support kit | Applied AI Foundations, the templates material | Build outreach, objection handling, FAQ answers, customer macros | AI Sales/Support Kit |
| 19 | Build a reusable AI assistant | OpenAI's creating-a-GPT guide or the Claude Projects docs | Build a no-code assistant grounded on your own docs and instructions, and use it daily (your first packaged skill) · ▶ worked example | Custom AI Assistant |
| 20 | Deep research as a skill | Your model's official guide: ChatGPT deep research, Gemini Deep Research, or Perplexity Research mode | Run a multi-step deep-research report, check every citation, and note where it is wrong | Deep-Research Report |
| 21 | 🛠 Apply day — Phase 3 capstone | No new learning | Build a mini campaign: article, carousel, video, three posts, one email | Mini AI Campaign |
Phase 4 — Systems: automation, agents, and proof (Week 4: Days 22–30)
By Day 30 you can: build a no-code automation, read and improve code with AI, ship a working prototype and a small feature, and build a real agent with a human approval gate — then evaluate it on known-answer tests, red-team it for prompt injection, and publish the case study. That last part is the portfolio that proves all of it.
Build it on $0. Every day this phase has a no-cost path, even though this is where "free" gets thinnest. For the code and agent days, use OpenAI Codex in the free ChatGPT tier and GitHub Copilot's free tier, get a free Gemini API key from Google AI Studio to power your agent, run any labs in free Kaggle or Colab notebooks, and host the finished demo on a free Hugging Face Space. Claude Code is the paid upgrade — the only one of the three with no free tier — and Anthropic's Claude Code courses are free even though the tool is not.
Pick your lane. This is where most 30-day plans lose their non-technical readers: one day is drag-and-drop automation and the next suddenly assumes you write code. You do not have to. Both lanes end at the same Day 26 artifact — a working agent with a trigger, tools, a loop, and a human approval gate — and Days 27 to 30 apply identically to both. If you have never programmed, run the no-code lane and skip nothing. (Taking the code lane? The [coding agents section](#know-your-coding-agents) below untangles Claude Code vs Codex vs Copilot before you start.)
| Day | Focus | Today's lesson | Build (45–120 min) | Artifact |
|---|---|---|---|---|
| 22 | No-code automation | Zapier's automation basics path | Both lanes: pick a high-volume, clear-rules, low-stakes task, then build a flow: form submission → AI summary → email draft → spreadsheet row · ▶ worked example | First Automation |
| 23 | Code literacy | Generative AI for Beginners (Microsoft), the code-generation lesson | No-code: map one real workflow as trigger → steps → decision points → output, then have ChatGPT or Claude critique the map. Code: have AI explain, improve, and document a simple script or page | Code Literacy Exercise |
| 24 | AI-assisted building | Claude Code 101 (free course, paid tool) or Copilot's quickstart. Stretch: Claude Code in Action | No-code: build a Custom GPT or Claude Project with instructions, your own docs, and one connector or Action. Code: write, test, debug, and document a small feature with Copilot and a coding agent · worked examples: Codex feature loop, Claude Code refactor, package a SKILL.md | Mini Build |
| 25 | Ship a prototype | The Google AI Studio quickstart | No-code: ship a prototype with AI Studio's app builder or Zapier Interfaces. Code: ship with AI Studio, Replit, Bolt, Lovable, or v0 — minding the credit meters · worked examples: app from a plain-English spec, or with Claude Code build a real website or a mobile app with Expo | AI Prototype |
| 26 | Build a real agent & the loop | No-code: Agents and Workflows (OpenAI Academy), first lessons. Code: Hugging Face Agents Course, Unit 1 — the longest single session this month, give it real time. Both: read Loops, explained first | No-code: build the agent in Zapier Agents, Make's AI agents, or an n8n template: trigger, context, tools, approval gate, final action. Code: build it with HF Unit 1 or Claude Code / Codex — same anatomy, same approval gate · ▶ worked example: the capstone agent, built | Working AI Agent |
| 27 | Evaluate it (evals for humans) | Evaluating AI Agents, the opening video | Test your agent and workflows on known-answer cases, do error analysis, and log the failure modes · ▶ worked example | Eval Notes |
| 28 | Agent & connector safety | Red Teaming LLM Applications, the opening video. Stretch: the MCP course to understand what your connectors actually do | Red-team your own tool: prompt injection, over-permissioned connectors, approval gates, and when to use restricted or lockdown modes · worked examples: red-team kit, connect via MCP | Safety Checklist |
| 29 | 🛠 Final case study + honest ROI | No new course — reread your Day 1 Opportunity Map first | Document one end-to-end workflow (problem, tools, process, outputs, risks, lessons) and baseline the task to measure the real time saved | AI-Native Case Study |
| 30 | 🛠 Apply day — Phase 4 capstone: public proof | Retake the [Day 0 self-assessment](#day-0-measure-your-baseline) and compare | Publish your artifacts and explain what you can now do with AI | Public AI-Native Portfolio |
Want a daily nudge while you run this? Agentic Daily sends one practical AI brief every weekday: what shipped, whether it's real, and what to do about it. Subscribe free and let the news do your Day 1 research for you.
Know your coding agents
Phase 4 leans on three coding tools that get conflated constantly. They are not the same thing.
Claude Code
Claude Code is Anthropic's agentic coding tool. It reads your codebase, edits files, and runs commands across the terminal, your IDE, the desktop app, and the browser. Learn it for codebase onboarding, bug fixes, tests, documentation, refactors, and feature work, and learn its permissioning and command execution so you understand what it is allowed to touch. Note that it has no free tier: it needs a paid Claude Pro or Max plan or metered API billing, so treat it as OPTIONAL PAID. Human review is not optional.
OpenAI Codex
Codex, once retired, is now OpenAI's coding agent, included in every ChatGPT tier (the free tier works at the lowest limits). Treat it as a software engineer you brief, for planning, building features, refactoring, reviewing code, and generating tests. The skill is writing a clear task, reading the diff it produces, checking the tests, protecting secrets, and never shipping generated code unreviewed.
GitHub Copilot
Copilot is the IDE-native assistant, and it has a genuine free tier (a monthly allowance of completions and chat, no card required). Learn autocomplete, Copilot Chat, code explanation, test generation, debugging, repository instructions, and its agentic mode. Its edge is that it meets developers where they already work.
The ten artifacts
Do not call this an official certification from any AI provider. Call it what it is: the AgenticDaily AI-Native Proof-of-Work Challenge. You complete it by shipping ten artifacts.
Every day produces something — thirty small outputs in all, one in the Artifact column of each phase table. Those are your reps. The ten below are the flagship pieces, pulled from specific days, that go on the wall: the portfolio you publish on Day 30. Don't polish all thirty; polish these ten.
*And here's the real payoff — every workflow you build is a skill, not a one-off.* A prompt you proved on Day 2, the meeting recipe from Day 10, the agent charter from Day 26 — each is a reusable capability you compose later, and the library compounds where a finished course just ends. That's literal, not just a metaphor: a Custom GPT or Claude Project (Day 19) packages a skill the no-code way, and a SKILL.md file — a folder your agent loads on demand — packages it the code way. Finish the month with ten prompts that work and two skills your agent applies unprompted, and you haven't completed a course; you've built a toolkit that keeps paying out.
- AI Opportunity Map (Day 1) — one page: 20 weekly tasks scored by frequency, time, and stakes, top 3 circled. Done when someone else could tell what you're automating this month.
- Prompt Library (from Day 2, compounds all month) — a living doc of your best prompts, each with role/context/task/format and when to use it. Done when you reach for it instead of rewriting from scratch.
- Model & Cost Scorecard (Day 3) — one task run across models and tiers, scored on quality, speed, and cost. Done when it tells you which model you'd pay for and why.
- Grounded Research Brief (Day 5) — a one-page memo where every claim carries a citation back to your own documents. Done when you can defend any line by clicking through.
- Responsible-AI & Data-Boundary Playbook (Day 6) — what you may and may not paste, per tool and tier, plus your bias, copyright, and disclosure rules. Done when a colleague could follow it without asking you.
- AI Chief of Staff Pack (Day 14) — your prompts, templates, and daily workflows packaged in one place (doc, Notion page, or folder). Done when tomorrow's work actually runs through it.
- Custom AI Assistant (Day 19) — a Custom GPT or Claude Project grounded on your own docs and instructions. Done when it answers a real work question better than a vanilla chat.
- Mini AI Campaign (Day 21) — an article, carousel, video, three posts, and one email, all from one research brief. Done when it's coherent enough to actually publish.
- Working AI Agent (Days 26–28) — trigger, context, tools, approval gate, final action, plus your eval notes and safety checklist. Done when it survives a known-answer test and your own prompt-injection attempt.
- Final AI-Native Case Study (Day 29) — problem, tools, process, outputs, risks, lessons, and the measured time saved. Done when a stranger could judge your capability from it alone.
The phase capstones (Days 7, 14, 21, 30) are your public checkpoints rather than separate flagships — Day 7's Personal AI Playbook, for instance, is where artifacts 2 and 5 get consolidated.
Keep yourself honest. Thirty days is exactly long enough to quietly quit. Download the free 30-day tracker — every day, lesson, build, and artifact pre-filled, with columns for your artifact link, one thing you verified, and minutes spent. Then make the capstones public: post your Day 7, 14, 21, and 30 artifacts with #AINative30. Public checkpoints are the accountability paid cohorts charge for, and your Day 0 and Day 30 self-assessment scores belong in that final post.
Score yourself
The challenge is built on six skills, each with proof attached — score each category by whether its artifacts exist and you could demo them to a colleague:
| Skill | Proof | Points |
|---|---|---|
| AI literacy, verification, and data governance | AI Opportunity Map, Verification Rubric, Responsible-AI Playbook | 25 |
| Prompting, model and cost selection | Prompt Library, Model & Cost Scorecard | 15 |
| Productivity workflows | AI Chief of Staff Pack, Grounded Research Brief, Custom AI Assistant | 15 |
| Creative and business workflows | Mini AI Campaign | 10 |
| Building, agents, evals, and safety | Working AI Agent with eval notes and safety checklist | 25 |
| Portfolio and public case study | Case Study, public portfolio | 10 |
| Total | 100 |
| Score | Badge |
|---|---|
| Below 70 | Not yet — rerun the phase you skimmed and re-score |
| 70–79 | AI-Native Explorer |
| 80–89 | AI-Native Builder |
| 90–100 | AI-Native Operator |
The course and tool directory
The day tables above tell you what to do each day — the "Today's lesson" cell points at the exact lesson or section for those 15 minutes. This directory is different. It exists so you can pick one deeper course to run alongside the whole month for real depth and a credential, and it doubles as the full link reference for every course, tool, and guide the plan names. You don't take everything here — you take the daily path plus one course from your level.
Pick one course to go deep. Choose the level that matches you, start its first course on Day 1, and chip away all month. Two columns do the deciding for you: In the plan shows which courses the daily schedule already sends you to (so you can tell a scheduled lesson from an optional swap), and You'll be able to says what you actually walk away with — pick on that, not on the certificate. The more technical your level, the more of this is genuinely extra. Time and free-tier notes are as of July 2, 2026; several providers gate graded work and certificates behind a paid membership even when the videos are free, so confirm what "free" includes before you count on a certificate. For the living version of this list, see our curated free courses.
Level 1: AI literacy for everyone
Best for non-technical professionals, executives, consultants, marketers, founders, and operators. Heads up: at this level the daily path already walks you through most of these, so the "go deep" course is often just finishing one you've started. The rows marked "alternative" are swaps if you want a different on-ramp.
| Course | In the plan | You'll be able to | Time | Cost & certificate |
|---|---|---|---|---|
| AI Foundations (OpenAI Academy) | Days 1–2, 6 | Use ChatGPT competently: prompt, add context, judge outputs | 60–75 min | FREE (ChatGPT account) · free cert (not OpenAI's formal Certification) |
| AI for Everyone (Coursera, Ng) | Day 1 (optional) | Scope an AI project and talk strategy with a technical team | ~7 hrs | FREEMIUM — free to audit (some Coursera courses now preview Module 1 only; specializations can't be audited) · cert ~$49 |
| AI Prompting for Everyone (DeepLearning.AI, Ng) | Day 2 | Write structured, reusable prompts instead of one-offs | ~3 hrs, 21 lessons | FREE during the DeepLearning.AI platform beta (VERIFY) · certificate needs Pro |
| Generative AI Fundamentals badge (Google Skills) | Day 10 | Explain LLMs and responsible AI basics, with a shareable badge | ~2 hrs (3 short courses) | FREE · free skill badge |
| Generative AI for Everyone (Coursera, Ng) | alternative | Spot generative-AI use cases in your own job | ~6 hrs | FREEMIUM — same audit caveat · cert paid |
| Elements of AI (Univ. of Helsinki) | alternative | Build solid AI literacy from zero, no code | ~30–50 hrs, self-paced | FREE · free cert |
| AI Fundamentals (IBM SkillsBuild) | alternative | Cover the same literacy ground with IBM's framing | ~6 hrs | FREE · free digital badge |
Andrew Ng's AI for Everyone is the strongest business-friendly pick: the vocabulary, how ML projects actually run, AI strategy, and the societal stuff you should not hand-wave.
Level 2: AI builder and workflow operator
Best for product managers, analysts, consultants, creators, founders, and semi-technical operators. This is the level where "go deep" pays off most: several of these (the Claude API course, Agentic AI, Microsoft's AI Agents for Beginners, and the Kaggle intensive) are real extra depth the daily path doesn't schedule. Rows marked "alternative" are optional extras beyond the plan.
| Course | In the plan | You'll be able to | Time | Cost & certificate |
|---|---|---|---|---|
| Anthropic Academy (hub) | the hub for the rows below | Work Claude end to end, from chat to API | Self-paced catalog, roughly 20 hrs | FREE · free cert per course |
| AI Fluency: Framework & Foundations (Anthropic) | Days 3–4 | Decide what to delegate to AI and how to verify what comes back | 1–2 hrs | FREE · free cert |
| Claude 101 (Anthropic) | Day 11 | Use Claude's projects, artifacts, and research modes on real work | 1–2 hrs | FREE · free cert |
| Applied AI Foundations (OpenAI Academy) | Days 8, 9, 18 | Turn one-off prompts into repeatable workflows | ~75 min | FREE (ChatGPT account) · free cert |
| Generative AI for Beginners (Microsoft) | Day 23 | Build generative-AI apps from working code samples | 21 lessons | FREE · no cert |
| Claude Code 101 (Anthropic) | Day 24 | Drive a coding agent through a real repo | ~1 hr | FREE course (hands-on needs a paid plan or API key) · free cert |
| Claude Code in Action (Anthropic) | Day 24 (stretch) | Wire the agent into MCP servers, hooks, and CI | ~1–2 hrs | FREE · free cert |
| Google AI Studio tutorials (Google) | Day 25 | Prototype with Gemini and get a free API key for your agent | Self-paced docs | FREE · no cert |
| Agents and Workflows (OpenAI Academy) | Day 26 (no-code) | Direct an agent through multi-step work with checkpoints | 75–90 min | FREE (ChatGPT account) · free cert |
| AI Agents Course (Hugging Face) | Day 26 (code) | Build a working agent and deploy it to a public Space (full course; Day 26 covers Unit 1) | 3–4 hrs per unit | FREE · free cert |
| Evaluating AI Agents (DeepLearning.AI + Arize) | Day 27 | Test an agent systematically and find its failure modes | ~2 hrs, 15 lessons | FREE during the DeepLearning.AI platform beta (VERIFY) · certificate needs Pro |
| Red Teaming LLM Applications (DeepLearning.AI + Giskard) | Day 28 | Attack your own app with prompt injection, then patch it | ~1.5 hrs | FREE during the DeepLearning.AI platform beta (VERIFY) · certificate needs Pro |
| MCP: Build Rich-Context AI Apps (DeepLearning.AI + Anthropic) | Day 28 (stretch) | Connect an agent to your real data with MCP | Short course | FREE during the DeepLearning.AI platform beta (VERIFY) · certificate needs Pro |
| Building with the Claude API (Anthropic) | alternative | Build production apps on the Claude API: tools, streaming, caching | 8+ hrs, 84 lessons | FREE · free cert |
| Agentic AI (DeepLearning.AI, Ng) | alternative | Build reflection, tool-use, planning, and multi-agent patterns in Python | ~8 hrs, 31 lessons | FREE during the DeepLearning.AI platform beta (VERIFY) · certificate needs Pro |
| AI Agents for Beginners (Microsoft) | alternative | Build agents on Microsoft's framework, lesson by lesson | 15 lessons | FREE · no cert |
| 5-Day Gen AI Intensive (Google & Kaggle) | alternative | Ship a capstone gen-AI project in Kaggle notebooks | 5 days, self-paced | FREE · Kaggle badge via capstone |
Level 3: technical AI foundation
Best for engineers, analysts, technical founders, and serious learners who want the math and mechanics under the tools. None of these are scheduled in the 30 days — every row here is a serious parallel commitment you run alongside the month, which is why the whole "In the plan" column reads "parallel track." Pick one.
| Course | In the plan | You'll be able to | Time | Cost & certificate |
|---|---|---|---|---|
| Machine Learning Specialization (DeepLearning.AI + Stanford, Ng) | parallel track | Train supervised models and neural networks properly | ~2–3 months | FREEMIUM — audit the individual courses free; the full specialization can't be audited · cert via Coursera/DeepLearning.AI Pro |
| 6.S191 Introduction to Deep Learning (MIT) | parallel track | Train deep networks in labs that now cover LLMs and agents | ~10 lectures + labs | FREE · no cert |
| Neural Networks: Zero to Hero (Andrej Karpathy) | parallel track | Build a working GPT from scratch, line by line | ~19 hrs of video | FREE · no cert |
| Practical Deep Learning for Coders (fast.ai) | parallel track | Train and deploy real deep-learning models from lesson one | ~20+ hrs | FREE · no cert |
| LLM Course (Hugging Face) | parallel track | Fine-tune transformers: tokenizers through training | Multi-chapter | FREE · free (per-chapter quizzes) |
| 6.036 Introduction to Machine Learning (MIT OCW) | parallel track | Work the ML theory: modeling, prediction, the math underneath | Self-paced | FREE · no cert |
| CS50's Introduction to AI with Python (Harvard) | parallel track | Ship 12 Python AI projects, from search to NLP | 7 weeks | FREE (edX verified cert is paid) · free CS50 cert |
| 6.034 Artificial Intelligence (MIT OCW, 2010) | parallel track | Understand classical AI — search, logic — as useful history | ~30 lectures | FREE · no cert |
| CS229: Machine Learning (Stanford) | parallel track | Follow graduate-level ML theory (2018 lectures + notes) | ~20 lectures (2018 set) | FREEMIUM — 2018 materials free; current term needs Stanford access · no cert |
Start with the Machine Learning Specialization, then pick a lane: 6.S191 or fast.ai if you want depth fast, Karpathy if you want to build a model with your own hands, the Hugging Face LLM Course if LLMs specifically are the goal. These assume some comfort with linear algebra and basic calculus; if you need that foundation first, Gilbert Strang's MIT 18.06 Linear Algebra is the standard free course.
The full tool stack
You do not need all of these on day one. This is the map of what AI-native work touches, so you know what each category is for when the roadmap sends you there. Pricing and free tiers move fast, so cross-check anything before you commit against our AI tools directory.
| Category | Tools to learn | What they're for |
|---|---|---|
| AI assistants | ChatGPT, Claude, Gemini | Writing, reasoning, analysis, brainstorming, document work, daily productivity |
| AI research | Perplexity, NotebookLM, Google Search | Source-backed research, current information, document synthesis, verification |
| Writing workspace | ChatGPT Projects & Canvas, Gemini Canvas, Claude Artifacts | Ongoing workspaces for writing, editing, and iterative creation |
| Design | Canva | Presentations, social graphics, brand assets, one-pagers |
| Video | HeyGen, Runway, Descript | Avatar videos, explainers, clips, captions, repurposing |
| Presentations | Gamma, Canva | AI-assisted decks, reports, microsites |
| Automation | Zapier, Make, n8n | Triggers, actions, filters, approval steps, agents |
| Coding assistants | GitHub Copilot, Cursor, Replit AI | IDE autocomplete, code explanation, test generation, pair programming |
| Agentic coding | Claude Code, OpenAI Codex | Codebase understanding, bug fixes, refactors, PRs, tests, multi-step coding |
| Prototyping | Google AI Studio, Replit, Bolt, Lovable, v0 | Fast prototypes, app mockups, chatbots, internal tools, landing pages |
| Learning | OpenAI Academy, Anthropic Academy, Google Skills, DeepLearning.AI, Hugging Face, MIT OCW | Fundamentals, responsible AI, agents, coding, ML literacy (all linked above) |
A few honest flags, because "free" is doing a lot of work in some of these categories: the video and prototyping tools (HeyGen, Runway, Gamma, Descript, Bolt, Lovable, v0, Cursor, Replit) have mostly moved to one-time credit buckets — great for a demo, thin for daily use. And n8n is fair-code, not classic open source: self-hosting is free for internal use, but hosting it as a service for others is not.
The responsible-AI checklist
Run this before you publish, send, deploy, or automate anything AI-generated:
- Do not paste confidential company data into tools unless you understand the data policy, and know the line between the consumer tier and an enterprise tier your org has cleared.
- Verify factual claims against sources.
- Review outputs for bias, tone, and context.
- Treat anything an AI reads (web pages, emails, documents, connector data) as a possible source of hidden instructions, and never give an agent more access than the task needs.
- Do not ship AI-generated code that handles payments, authentication, health data, customer records, or sensitive processes without technical review.
- Label or disclose synthetic media where appropriate.
- Check copyright, licensing, and brand usage.
- Keep a human approval step for high-impact decisions.
- Use AI to accelerate judgment, not to replace it.
Do the work
Do not spend the next 30 days reading about AI. Spend them building with it. Take the free courses, use the freemium tools, make the ten artifacts, and publish the proof. That is the whole difference between people who talk about AI and people who are AI-native, and it is available to you at no cost starting today.
Start Day 1 now, and subscribe to Agentic Daily to get one practical AI brief every weekday while you work through the plan.
Agentic Daily — daily AI intelligence for people who build.
This challenge is educational and is not legal, medical, financial, or investment advice. If your work involves patient data, privileged or client matters, or regulated financial information, follow your organization's compliance, privacy, and HIPAA or equivalent requirements, and confirm a tool's data policy, before putting any real data into it. Keep a human review step on anything high-impact.
Source notes
Pricing, free tiers, credits, and certificate rules change often, so verify before you rely on any of these. Every link, free-tier claim, and certificate claim in this article was last verified on July 2, 2026. A handful of help-center links (Canva, Descript, OpenAI's GPT and deep-research guides, Perplexity, Zapier) block automated checkers, so those were verified by hand. If you find something that has gone stale, reply to any Agentic Daily issue and we will fix it.
Changelog. July 3, 2026 (v4): added ▶ worked-example links on the build days — a companion pack of copy-paste workflow recipes (prompt rewrite, email bank, meeting notes, personal assistant, first automation, the capstone agent, Codex feature loop, eval harness, red-team kit, MCP connect, packaging prompts as SKILL.md skills, and building a real website and an Expo mobile app with Claude Code); added the "every workflow is a skill" framing. July 2, 2026 (v3): restructured into four phases (one row per day, each pairing the day's lesson with its own direct course link instead of a shared academy-homepage link); added the course and tool directory, the level router, and the MCP and Claude Code in Action stretch courses; replaced an unsourced productivity statistic with the Danish adoption study and METR trial figures; corrected course durations (AI Prompting ~3 hrs, AI for Everyone ~7 hrs, Microsoft AI Agents 15 lessons) and the Coursera audit caveat. July 1, 2026: added the Hugging Face, Microsoft, and Google/Kaggle agent courses; refreshed the Level 3 technical track; added the $0 build stack, the Day 0/30 self-assessment, the day-by-day learn schedule, the Week 4 no-code lane, and a changelog. Original publication: June 2026.
- Anthropic AI Fluency framework: anthropic.com
- OpenAI Academy: courses and certificate clarification
- Anthropic Academy: courses
- Hugging Face: AI Agents Course and LLM Course
- Microsoft: AI Agents for Beginners and Generative AI for Beginners
- Google and Kaggle: 5-Day Gen AI Intensive
- MIT 6.S191 Introduction to Deep Learning, Andrej Karpathy's Neural Networks: Zero to Hero, and fast.ai
- Sensitive data in consumer AI tools: Cyberhaven 2026 AI Adoption & Risk Report
- Measured vs perceived AI productivity: Humlum & Vestergaard, "The Adoption of ChatGPT" (25,000 Danish workers, ~3% average time savings) and METR's 2025 developer RCT (19% slower with AI while believing ~20% faster)
- Claude Code plans: pricing and access
- OpenAI Codex: codex
- GitHub Copilot: plans
- Andrew Ng, DeepLearning.AI: AI for Everyone, Machine Learning Specialization, Agentic AI, and Red Teaming LLM Applications
- Google Skills, Generative AI Fundamentals: skills.google
- MIT OpenCourseWare: 6.034 AI and 6.036 ML
- Harvard CS50 AI: course site
- Stanford CS229: course site
Get the next one in your inbox