mainiø 1.1.4: onboarding that proves itself, full iPad, and a smarter progress ring

Onboarding used to ask for a signup on faith. It now proves the point first: you build an improved prompt by hand, see it work, and get a starting learning plan shaped around what you told us you want from AI, all before we ever ask you to sign in.

Sign-in comes after that proof, not before it. Returning users also get a fast sign-in escape hatch instead of being pushed through the whole flow again, and we only ask for a name when we cannot infer one quietly, so there are fewer forced fields between someone and their first real lesson.

iPad support is live, and it was worth doing right rather than fast. An earlier build briefly had it, but a device-family issue meant that release shipped iPhone-only while we fixed it properly rather than ship something broken.

Layouts are actually rebuilt for the bigger screen, not an iPhone view stretched wider. It is one universal download and one subscription across iPhone and iPad, so anyone using mainiø on iPad gets the real thing, not a placeholder.

The home ring used to show one number: how much of your path you had completed. It now draws two arcs on the same ring, a lighter one for completed and a darker one on top for completed with a full score, so you can tell "I did this" from "I actually got this right" at a glance.

The same release fixed the object-removal lesson properly: image generation no longer fails, all three practice photos can be chosen, the painting area is far larger, and a new full-screen mode makes it clear what you are actually erasing. Image styles no longer come back as mostly empty pictures either.

Underneath, the paid content got deeper too. The Tool Landscape path for mainiø+ grew from one lesson to eight, with new quizzes and flashcards on actually running a tool stack: lock-in, testing something new, buy versus build, and what to do when your stack breaks.

The first Academy course teaches the basics: what AI can do, how to prompt it, and how to think critically about the result. Course two, From Prompt to Workflow, picks up where that leaves off. It is for people who already write reasonable prompts but want the output to be something they can actually review, reuse, and repeat.

It runs across four chapters and twelve lessons: defining the job before you prompt, making prompts reusable with templates and maintained context, improving a weak answer instead of just rewriting it, and building workflows with review gates and a way to debug when something goes wrong.

Alongside it, the original course gained a new foundations lesson, How AI Fields Fit Together, for the very common confusion between AI, machine learning, data science, generative AI, and agents. Between the two courses, Academy is now at 37 lessons, still free, still no account required.

Apple lets apps run in-app events that show up as a card on the App Store before someone has even installed the app. We built the web side of that: dedicated event landing pages, an apple-app-site-association file, and the routing so a tap on the event card can universal-link straight into the matching mission inside mainiø, not just to the app's home screen.

The first event, Visualise Your Next Idea, sends people directly into the image-creation lesson where they turn a written idea into a real generated image. If mainiø is already installed, the link opens the exact lesson. If it is not, the landing page explains the same five-minute exercise and links to the App Store.

It is a small piece of plumbing, but it is the difference between "download an app and find your way around" and "try this one thing right now."

Academy analytics were minimal from the start, but "minimal" is not the same as "clearly explained." We rebuilt the footer analytics choice into a proper consent banner: Essential only keeps just the local progress-saving storage the course needs, and Allow anonymous analytics is what turns on anything else.

With consent, we record structured events — a page, chapter, or lesson opened, a lesson or practice mission completed — plus lightweight engagement signals like active reading time, interaction count, and scroll depth, so we can tell a lesson that is skimmed from one that is actually read. We do not record what anyone types, clicks, or answers, and there is no persistent identifier, only a random session id scoped to one browser tab.

We also wrote down our own rules for campaign links, so social and organic traffic gets attributed consistently without ever adding tracking parameters to links inside mainio.co itself. Academy event data is kept for at most 13 months and used for exactly one thing: understanding whether lessons are working.

For now

Onboarding and iPad were the two people had been waiting on, so they lead. But the rest matters just as much: a home ring you can trust, a deeper second course for people who want more than better prompts, a shorter path in from the App Store, and data we can look at without flinching.

The live-AI features get the headlines. This is the work that makes the next round of them worth trusting.