Mikey builds Lumen, a liquid-glass journal app, in six Gemini 4 Argon prompts on Base44: tags and search, AI reflection, sign-up, Stripe premium and a final polish.
Turn a vague app idea into a Base44 client CRM in 24 hours
Mikey No Code interviews a subscriber about her workflow, keeps 4 features and parks 5, feeds Base44 one complete prompt, then fixes real-use problems with her screenshotted feedback.
Mikey No Code · YouTube · published Oct 20, 2026 · Work & Operations
▶ Watch on YouTubePrompt
Build a simple client tracker for [business name], a one-person [type of business] in [city]. I'm the only user: no team members, roles or extra accounts. It replaces WhatsApp, a paper notebook and an old spreadsheet and must answer one question: who is waiting on me? Client records: full name, phone, email, address, inquiry source ([sources]), date the inquiry arrived, free notes. Pipeline stages: [stage 1 to 5]. Moving a client between stages must take one action. Each client has a quote value in [currency] and a next follow-up date; anyone past the date appears in a 'needs chasing' list at the top, showing name, days overdue and a quick way to open them. Each client has a dated activity log, newest first. Screens: board, client record, add-client form only. No settings, reports, analytics or login. Visual direction: calm and warm, like an independent stationery shop: off-white background, soft near-black text, one deep green accent. Avoid purple or blue gradients, dark mode, glassy cards, oversized rounded buttons, heavy shadows and stock illustrations of people. Fill the first build with 12 realistic example clients spread across all stages.
Visual breakdown
Mikey helps a subscriber with no spec or wireframe get a working client-tracking app in 24 hours. The key move: interview her about how her business runs, keep only 4 features, park 5, give Base44 one complete prompt, then fix what she finds in real use.
The problem: A business owner tracks inquiries across WhatsApp, a notebook and an old spreadsheet, had lost two jobs by not following up, and doesn't know where to start with an app idea.
From a vague idea to a live app
- Community postA subscriber has an app idea but no spec, wireframe or technical plan
- Workflow interviewWalk through the business from first inquiry to finished job, including the annoying parts and one recent problemDropped: Feature wish-list thinking
- Scope decisionsShe picks what she'd miss first on Monday morning: 4 features go in, 5 are parkedDropped: Invoicing, auto emails, client portal, calendar sync, multiple logins
- One complete promptBusiness context, records, 5 pipeline stages, screens, visual direction, avoid list and 12 sample clients go into Base44 at once
- Feedback roundsBase44's open questions go to her; her screenshotted replies go back in discuss mode before changes are applied
- Live appRenamed URL, works on her phone, security scan run before handing it over
Starting pointHuman workBase44Result
What made version one
- Client records: name, phone, email, address, inquiry source, date, free notes
- A board with her own 5 stages: new inquiry, consultation booked, quote sent, project live, completed
- Quote value in pounds and a next follow-up date, with overdue clients in a 'needs chasing' list at the top
- A dated notes log inside each client
By the numbers
24 hoursidea to working app
4features built first
5features parked
12sample clients in the first build
Do it for your own client or idea
The thinking comes first; the Base44 prompt is the easy part because every line comes from something the user already said.
- Ask how the business runs day to day, from first inquiry to finished job, who uses the app, and one recent problem they wish hadn't happened.
- Get one sentence: what the app should keep in one place and why it matters. Use it as the rule for every feature.
- Ask which thing they'd miss first on Monday morning to rank features, and park the rest with a reason.
- Translate the feel into a visual direction plus an avoid list, then put everything into Base44 as one complete prompt with sample data.
- Ask Base44 which 3 or 4 decisions it had to guess at, turn them into plain questions and send them to the user.
- Screenshot their answers into Base44 in discuss mode, agree the plan first, then apply it. Let them use the app on a real workday and repeat once.
- Rename the free app URL under domains, run the security scan, and hand it over.
The video shows a reusable version of the prompt with placeholders for business name, city, stages, sources, currency, accent colour and sample data.
Lessons from the build
- When building is fast, deciding what not to build is the hardest part. Every parked feature had a reason.
- The first request is often a symptom: reminders wouldn't have fixed a problem caused by checking three places.
- Give Base44 the full picture at the start rather than five small prompts that keep changing the app's shape.
- Real use finds what a demo doesn't: follow-up date wasn't required, so a client could vanish from the chasing list.
- Describing what you don't want is one of the most reliable ways to stop an AI-built app falling back to the default look.
Jump to the part you need
0:00The challenge: a rough idea, no spec2:34The subscriber's post and why scope matters4:15Interview about how her business runs7:04Choosing 4 features to build first8:19Five features parked and why9:38Her five stages and the visual direction12:05The one complete Base44 prompt16:01Publish version one and ask her questions18:08Applying her answers in discuss mode20:36Problems found in real use25:34Rename the URL and hand it over27:13Security scan and wrap-up
Distilled from the video’s captions
How to try it
- Open Base44.
- Copy the prompt and fill in the [placeholders].
- Watch Mikey No Code's video for the full walkthrough and setup.
Related use cases
Skill Leap AI builds two Gemini Skills: an Executive Summary skill that condenses any long document to one page, and a Data Snapshot skill that turns a spreadsheet into a short summary plus a chart, with no prompt needed.
OpenAI shows team tasks in ChatGPT: set up recurring cloud-run work the team manages together, using a legal team's daily intake of requests and deadline checks, then refine the shared instructions for the next run.
Nate's agency runs humans plus their own Claude Code and Codex subscriptions in one Kylon workspace, so agents and skills help the whole team, not one person. Sponsored.