How to learn AI the right way and skip the LinkedIn doomscroll trap
Every morning a new 'AI destroyed [profession]' post crosses your feed. None of them are right. The calculator analogy, the kindergarten test, and the 90-day curriculum that actually moves you.
Summarize this article with:
- The calculator test. Calculators didn't bankrupt mathematicians. They made them 100x faster. AI follows the same pattern for every profession that doesn't replace itself.
- The kindergarten rule. Give a calculator to a 5-year-old and you ruin their ability to do mental math. Same risk with AI: junior practitioners who skip the fundamentals.
- The Foxpro/Flash/Dreamweaver schedule. Tools die on a predictable cycle. Skills transfer. The list of what survives.
- The 90-day AI learning curriculum: Week 1-2 (fundamentals), Week 3-6 (specific tool depth), Week 7-10 (your domain x AI), Week 11-12 (build something real).
- The 5 signs you're on the right learning path vs the LinkedIn-doomscroll path.
Most 'how to learn AI' content is marketing for the person writing it. They want you to feel behind so they can sell you their cohort. The honest version is shorter, slower, and produces more actual capability. This is that version.
Every morning, my LinkedIn feed serves me 3 to 5 versions of the same post.
"AI just destroyed SEO. Here is what I am giving you for free."
"Claude killed software engineering. Why I am hiring zero junior devs."
"AI destroyed marketing. The only skill that matters now is prompting."
None of these posts are true. The people posting them know they are not true. They post them because doom is engagement-bait, and engagement-bait sells their cohort, their course, their consulting practice.
This piece is the opposite. The honest version of how to learn AI. It is less exciting. It will not get you 10x leads in 30 days. It will get you actually capable at AI inside 90 days, which is the only outcome that matters.
Start with the calculator test
When the calculator went mainstream in the 1970s, the predictions were identical to today's AI predictions. Mathematicians would be unemployed. Accountants would be obsolete. Bookkeepers would have nothing to do. Reality was different. The BLS Occupational Outlook Handbook counts roughly 1.6 million accountants and auditors in 2024, and the profession is still projected to grow through 2034. Mathematician-related occupations grew sharply too. The calculator did not end them; it amplified them.
What actually happened: the calculator made them 100x faster. The work that took 8 hours now took 5 minutes. The profession absorbed the productivity gain. Some jobs at the bottom of the skill curve disappeared (manual ledger-keepers, slide-rule mathematicians). New jobs at the top opened up (financial modelers, computational mathematicians).
AI follows the same pattern. The work that took 8 hours now takes 5 minutes. The profession absorbs the productivity gain. Some jobs disappear (junior copywriters who only write generic copy). New jobs open up (people who prompt AI well, edit it sharply, ship 10x more output than before).
The kindergarten rule
Here is the part the calculator analogy gets right that the LinkedIn-doomscroll posts miss: when you give a calculator to a kindergartener, you ruin their ability to do mental math forever.
AI follows the same pattern. Junior practitioners who skip the fundamentals and prompt their way through tasks for 2 years lose the ability to do the underlying work. They become unable to spot when AI is confidently wrong. They cannot debug. They cannot edit.
This is the actual risk. Not that AI replaces you. That YOU become unable to do the work without AI, in a profession that still requires the underlying skill 20% of the time.
The right time to use AI heavily is after you have the fundamentals. The right time for a junior to learn the craft is before they reach for AI as a default.
A senior with AI is 10x faster than a senior without it. A junior with AI is 0.5x as effective as a junior without it. The difference is fundamentals.
Tools die. Skills transfer.
Read this paragraph if you are worried about being replaced.
Microsoft announced FoxPro end-of-life in 2007 (Visual FoxPro 9.0 lifecycle); shops migrated to .NET through 2010-2015. Adobe announced Flash end-of-life in 2017; the platform died December 31, 2020. Adobe put Dreamweaver in maintenance mode in 2020 (still sold, no new features). Photoshop-only retouchers started competing with generative AI workflows in 2022. Each of these waves was real. Each one looked terminal at the time. Each one transferred to the next platform.
The FoxPro people learned Visual Basic. The Flash people learned HTML5 and CSS3 animations. The Dreamweaver people learned React. The Photoshop-only retouchers learned generative AI workflows. None of them ended up unemployed unless they refused to learn the next tool.
AI is the next tool. Refuse to learn it and you join the FoxPro holdouts. Learn it and you join the people who absorbed every prior wave.
The 90-day AI learning curriculum
Three months. Four phases. No cohort, no course, no $5K certificate.
Weeks 1-2. Fundamentals
Spend 30 minutes a day, every day. Use one LLM (Claude or ChatGPT, pick one) for every task you can. Writing emails. Drafting documents. Researching anything. Outlining slides. Coding helper queries.
The goal: feel the texture of where AI is useful and where it fails. Two weeks of constant low-stakes use teaches you more than reading 50 articles about prompting techniques.
What you measure: how often does AI's first answer save you time vs cost you time? After 2 weeks, you have a working internal model.
Weeks 3-6. Specific tool depth
Pick ONE tool to go deep on. The right pick depends on your work:
- Marketers, writers, founders: Claude or ChatGPT.
- Designers: Midjourney + Gemini (for image-to-image work).
- Developers: Cursor + Claude (Sonnet for editing, Opus for hard problems).
- Analysts: ChatGPT Code Interpreter + Gemini for big CSVs.
Read the official docs of that tool from start to finish once. Read 3 best-prompt-engineering articles. Then practice 30 minutes a day on real work.
What you measure: at the end of week 6, can you produce work in 1 hour that used to take you 3 hours? If yes, the tool clicked. If no, you picked the wrong tool. Restart with a different one.
Weeks 7-10. Your domain x AI
Apply the tool to your specific domain in a structured way. Not "use AI for marketing" but "rebuild my entire ad-creative pipeline using Claude + Midjourney."
Take one workflow you currently run manually. Map every step. Replace each step with AI where it makes sense. Keep the human-judgment steps human. Test the new workflow on real work for 4 weeks.
What you measure: hours saved per week. Output quality (better, same, or worse). Cost saved (subscriptions vs in-house). At the end, decide whether to make the new workflow permanent.
Weeks 11-12. Build something real
The capstone. Ship a real piece of work that you could not have shipped before AI. Not a demo. Not a tutorial. A real output your business or your client uses.
For marketers: a 30-variant ad campaign that you ran on real budget.
For founders: an audit of your own business that surfaces 3 specific fixes.
For developers: a small tool you built solo that would have needed a 2-person team before.
For designers: a brand-system rebuild that takes a week instead of a month.
The capstone matters because shipping something real proves to yourself you are capable. The capability is what you carry forward when the next AI wave hits in 18 months.
5 signs you're on the right path
- You can name 3 things AI does badly that you used to think it did well. (Means you have calibrated expectations.)
- Your output volume is up but your editing time is also up. (Means you are using AI without trusting it blindly.)
- You can fix AI's first answer in 2 minutes instead of accepting it or restarting. (Means the fundamentals are intact.)
- You stopped reading LinkedIn posts about AI 30 days in and just started using it. (Means you skipped the doom-content trap.)
- You can explain to a non-technical person, in 2 minutes, what AI is good for and what it isn't. (Means you have a working internal model.)
5 signs you're on the wrong path
- You spend more time reading "AI destroyed X" posts than using AI.
- You can describe 10 AI tools but use none of them daily.
- Your output is up but the quality is down and you cannot tell.
- You skipped the fundamentals because AI made them feel optional.
- You signed up for a $5K cohort and are 6 weeks in without shipping anything real.
Frequently asked questions
Should I take a paid AI cohort instead of the DIY 90-day curriculum?
Which LLM should I learn first: Claude, ChatGPT, or Gemini?
Is it too late to start learning AI in 2026?
What about prompt engineering courses?
Will AI replace my job in 5 years?
What's the one piece of advice you'd give a founder learning AI?
Related reads
- The Foxpro lesson. The longer-form history of why tools die and skills transfer.
- Which parts of your job to hand over to AI. The plumber rule for what to automate vs not.
- How to run a digital marketing audit using AI. Applied capstone for marketers.
Maddy
Maddy runs every WeActive8 engagement personally. Nine years working on growth across SMB and funded-startup stacks. Builds the 8CRM, Team8s, 8Host, and 8Automations products.
