The Offline Apps Experiment

01 — Where it started

It started with an irritation. Every useful little app I tried wanted an account, a monthly fee, and my data on someone else’s server. To track subscriptions, I was asked to take out another subscription.

I write about the Human Edge: the idea that AI can generate almost anything, but a human still has to decide what is worth making. At some point writing about it was not enough. I wanted to test it with my own hands.

So I wrote down a hypothesis: one person, with AI as a building partner, can ship everyday tools that solve one real problem each, run offline, keep data on the device, and are paid for once. Three rules, no exceptions: useful, private, fair.

02 — How I build: the machine drafts, I decide

I use the same split as Phoenix Soulfire™. The machine pass comes first, the human pass second.

AI drafts: code, interface variants, translations into English, Italian and German, first versions of product copy. It is fast, fluent and, left alone, interchangeable.

I decide: which problems deserve an app, what to cut, that everything works offline, that nobody needs an account, that the price is paid once, and whether a result is good enough to carry my name. A model can propose a feature. It cannot decide that a tool should respect the person using it.

Each app ships as a progressive web app plus a standalone .ZIP/.HTML package, so it opens in a browser and keeps working offline on iPhone, Android, Mac and Windows.

The sales-bot network

Then I hit the second half of the problem: building is not selling. Nobody finds a store by accident. So I extended the experiment to distribution and gave the store to a small network of AI agents, each with one job, coordinated by a primary bot. The network is part of what I am testing.

1. Grok Bot (primary) — routes tasks, coordinates the other bots and runs a "Regia" check three times a day.
2. Traffic Doctor — diagnoses why the store gets no traffic and fixes what it finds.
3. SEO & Organic Funnel — blog SEO, Google indexing and lead magnets.
4. AI Search Visibility — checks whether AI assistants and Google recommend the store.
5. CRO & Bundle Closer — conversion, bundles, discounts and product pages.
6. TikTok & Viral Traffic — short-form content for TikTok, Threads and X.
7. Pino · Social & Pinterest Growth — Pinterest and social presence.
8. Shopify bot — day-to-day store operations.

Human rules in the loop. Free methods only. No spam. Reversible changes only. Nothing is published automatically without my approval. Legal and fiscal pages are never touched by a bot.

The bots propose and execute; I keep the final word. Whether this network can bring real people to a small store without spam is exactly the question I do not know the answer to yet.

03 — The artifacts

Here is everything the experiment has produced so far: every active product in the store, grouped by what it is. Prices are as listed on bjbeyond.studio.

A · Money and daily life — the core four

A1Subscription Tracker Pro

Problem: Subscriptions you forgot keep renewing on your card.

What it does: Puts every subscription in one place, shows the real cost per use, writes the cancellation notice for you, adds renewal alerts to your calendar (.ICS) and projects five years of savings.

Price: €9.90, one-time. Store page · Live demo

A2Freelance Rate & Quote Pro

Problem: Freelancers undercharge because the hourly rate they quote ignores taxes and costs.

What it does: Works out the rate you actually need from an income goal, costs and billable hours (US/UK/EU tax presets), then turns it into a three-tier quote you can print as PDF.

Price: €14.90, one-time. Store page · Live demo

A3Fair Split Pro

Problem: Shared rent, bills and groceries turn into arguments about who owes what.

What it does: Splits costs 50/50 or by income, keeps a running balance of who owes whom, and shares the summary via WhatsApp. No account for either partner.

Price: €11.90, one-time. Store page · Live demo

A4Deadline Radar Pro

Problem: Insurance, car inspection, ID cards and warranties expire quietly, and a missed date costs a fine.

What it does: Puts every deadline on one urgency radar (overdue, due within 60 days, estimated cost for the next 90), with presets for car, home and documents, calendar reminders and an inspection and road-tax date calculator.

Price: €9.90, one-time. Store page · Live demo

B · Other tools from the same bench

B1Mind Map Pro

Problem: Ideas stay scattered in linear notes.

What it does: A mind-mapping web app, installable as a PWA, with local saving, exports and an English, Italian and German interface.

Price: €6.30, one-time. Store page

B2Synced Signal for Two

Problem: Couples want a quiet channel away from social feeds.

What it does: A private two-phone note signal for partners, running in any modern browser, with no recurring fee.

Price: €4.20, one-time. Store page

B3Wedding Countdown Timer

Problem: A wedding date deserves more than a calendar entry.

What it does: A personalised countdown with your names and date, as a standalone HTML app that runs locally in the browser.

Price: €6.30, one-time. Store page

B4Halloween Wallpaper Generator — Liquid Glass Studio

Problem: Ready-made wallpapers all look the same.

What it does: A desktop HTML generator (Pumpkin, Ghost, Moon themes; colour, glow, glass and stardust controls) that exports 1080 × 1920 PNGs offline. The artwork is procedural, not AI-generated.

Price: €30.00, one-time. Store page

C · Visual packs

Side-products of the same studio: wallpaper packs for phones and screens.

• Free Liquid Glass iPhone Wallpaper — Free
• 10 Liquid Glass iPhone Lock Screens — €2.10
• 10 Neon iPhone Live Wallpapers (HEIC & MOV) — €2.10
• Digital Detox iPhone Wallpapers + ambient audio — €10.50
• Halloween iPhone Wallpapers — 10 Liquid Glass designs — €5.25
• Halloween Liquid Glass Wallpapers & Digital Screens — €14.70
• Halloween Liquid Glass Bundle — Generator + 10 designs — €15.75

Why offline, why once

Money, deadlines and shared bills are private. A tool that handles them should not need a login or send them anywhere. And a tool that helps you cancel forgotten subscriptions should not become one. That became the design rule for everything above.

04 — Results so far

Honestly: early. The artifacts exist, they are live, and they work offline as designed. That part of the hypothesis holds.

The harder part, whether people find them useful enough to pay for, is still open. I will not publish numbers I do not have, and I will not dress up a beginning as a result.

05 — What happens next

I will listen to the people who use the apps, fix what they find, and keep the same three rules for anything new. The bot network keeps working under the same human limits.

This log will be updated as the experiment produces evidence, good or bad. If you try one of the tools, tell me what broke.