If you've ever started a course, a playlist or a "30 days to learn X" plan and quietly stopped halfway, this post is for you. It doesn't matter if X was React, UI design, Spanish, data analysis or exam prep. I spent a while reading what learners write about this, including students, self-taught developers and people switching careers. The same handful of problems kept coming up. I'm building Navigator, an AI learning roadmap app, so I wrote the problems down and checked whether my product answers each one. Here's the list, and what I did about it.
You finish one course, have no idea what comes next, and start another from zero. Watch-later lists with hundreds of videos, 73 bookmarks, ten open tabs, and nothing finished. Developers call it tutorial hell, but students and career switchers live in the same loop.
What I did: Navigator turns "I want to learn X" into phases. Every step is tagged Watch, Read, Build or Practice, so you can't spend three weeks only watching. There's also a Sprint mode (1 to 4 weeks, one goal, day-by-day tasks) for when you need a result fast, like a portfolio piece or an interview.
You follow along, everything makes sense, and then you try it alone and nothing comes out. Students know this as re-reading your notes and feeling ready for the exam, then blanking on the paper.
What I did: A Watch step isn't complete when you tick a box. It's complete when you pass a short quiz on what you just watched, and the next step stays locked until you do. I'm not sure this is right for everyone (see the questions at the bottom).
Someone told me on a forum that if they miss the finish date by a week, the roadmap starts feeling like a debt, so they stop opening the app. That one stuck with me, because I think it's why a lot of people quit. It's not that the material was too hard, it's that coming back feels bad.
What I did:
Streaks have freezes and grace days, so one missed evening doesn't wipe six weeks of work.
There's no fixed deadline. Navigator projects a finish date from your real pace and updates it as you go, so falling behind moves the date instead of turning the plan red.
Plenty of posts say getting good at something hard on your own often takes much longer than people plan for, especially part-time next to school or a job. Most learners plan for a few weeks, miss it, and decide they're not cut out for it.
What I did: The projected finish date is based on how fast you actually complete steps, not how fast you hoped to. There's also a free Learning Timeline Calculator (no sign-up) where you split a skill into parts and see a realistic date.
Isolation shows up on almost every list of reasons people quit. Nobody notices when you go quiet.
What I did: Streak partners. You pair up with a friend and see each other's daily plan, with lightweight reactions (a fire, a clap, a nudge) instead of long messages.
What I did: Navi, an AI tutor that already knows which step and topic you're on, so you're not pasting context into a generic chatbot. You can keep one conversation per topic and save any answer to your notes.
I also worry about the other side of this. A lot of recent posts point out that learners who let AI do the thinking end up with answers they can't explain or fix when something goes wrong. That's part of why the quizzes exist: the app should check that you understood, not just that you finished.
Every AI roadmap tool has the same obvious problem: the model confidently links to a video or article that doesn't exist.
What I did: The generator isn't allowed to invent URLs or sources. Every resource has to resolve to a real, reachable page or it's removed before it reaches you, and if a link breaks later the roadmap flags it.
"There are already AI roadmap generators"
Yes, and a few of them are on DEV. I looked at them before building more. The ones I found mostly generate a list of topics with links. I'm trying to compete on what happens after the list exists: verified links, checkpoints, a streak that forgives you, and a finish date that tells the truth.
If you've used one of those and I've got it wrong, tell me. Behind the scenes
For the builders reading: the part that surprised me most was keeping the AI side alive. Providers change plans, retire endpoints and rate-limit you, and my app broke more than once because one of them moved. I ended up with a multi-provider fallback chain behind a single edge function (React, TypeScript, Supabase), so no single provider going down takes the app with it.
Try it
Navigator is on a waitlist while I finish testing, but two tools are open with no sign-up:
Skill Crossover: pick a skill you already know and see what to learn next
Learning Timeline Calculator: break a skill into parts and get a finish date
Waitlist and both tools: https://navigator-web.netlify.app Questions for you
I'd rather hear honest opinions than compliments:
Quitting: what actually made you drop your last course, tutorial or study plan?
Streaks: do they keep you going, or make you feel guilty when you miss one?
Locked steps: should a step stay locked until you pass a quick quiz, or is that annoying?
Trust: what would make you trust a roadmap an AI generated?
Students, self-taught devs, career switchers, all welcome. Roast it if you want, that's more useful to me than polite.