Workshop-scale app ideas for laymen

TL;DR. For a 2-3 hour layman vibe-coding workshop, lock the scope shape to single user, single screen, no auth, localStorage-or-nothing persistence — the “rectangle of stuff on a page” pattern [1] [86]. From that shape, five reliable menus stack roughly by technical depth: (1) calculators and personal trackers (tip, BMI, habit, expense splitter) [12] [13], (2) decision toys and generators (random-restaurant wheel, name generator, Wordle clone) [19] [35], (3) fetch-and-render with one no-key API (Open-Meteo weather card, PokéAPI Pokédex, ISS tracker) [44] [46] [51], (4) scratch-an-itch family or teacher utilities (meal planner, lesson critique, baby-sleep dashboard) [76] [78] [80], and (5) micro-games (Pong, Hangman, falling-objects catcher) [25] [36]. The categories that consistently blow the timebox: multi-user marketplaces, real-time multiplayer, anything with payments, native mobile App Store deploys, and OAuth-provider integrations [2] [87] [90] [91] — when an attendee proposes one, swap it for the single-user variant.

What “workshop-scale” actually means

Instructors converge on a sharp shape. Stanford’s non-developer Lovable course caps four modules’ output at a to-do list or feedback form [8]. Dyad’s beginner guide names todos and quizzes as the sweet spot — “small enough to finish in an afternoon, complex enough to teach real patterns” — and even pins concrete UI shape: 4 multiple-choice answers, a score counter, a 30-second countdown [12]. At Singapore’s 65labs, instructor Sherry Jiang puts it bluntly: “vibe coding is best suited to very simple, lightweight, straightforward consumer apps — not anything very technical” [4]. The Innovation Co-Lab at Duke ran the same format in February 2026 for the same audience [70].

The constraint table laymen should hold their idea against:

The “last 10% is 90%” effect is real. Calvin Ku reached 90% on a “simple” Pomodoro in three days and then spent months on the last 10% once OAuth, accessibility, and DB wiring entered scope — his post-mortem rule: one primary loop and one settings screen, defer everything else [6]. Over 60% of security issues in vibe-coded apps come from the moment login enters the picture — missing row-level security, exposed API keys, client-side-only auth, IDOR — and beginners cannot debug any of these live [10] [92]. Tool choice should match scope too: v0 or Tempo for a single component or landing page; Lovable, Bolt, or Replit only when stored data is actually needed [5]. Five minutes of up-front “one paragraph + screens list” documentation saves hours of redirects later [7].

Menu 1 — Calculators and personal trackers

The safest first project. Each fits one screen, ≤3 inputs, and either localStorage or no persistence at all.

The tip calculator [13], Ann Jose’s themed Pomodoro Flow [22], Questera’s habit-tracker template [71], and the four Lovable-deployed calculators [15] [16] [17] [24] all demonstrate the same structural pattern: one input form, one output panel, one optional localStorage list. Event Countdown [18] and the Devpost expense splitter [23] push the high end of “still fits in 2-3 hours” — useful when an attendee wants a slightly meatier challenge. Taskade’s 2026 calculator roundup adds Meeting Cost Calc, QR Code Studio, Color Palette Extractor and Multiplication Playground in the same scope band [14].

Menu 2 — Decision toys, generators, and word games

Higher fun-per-prompt ratio than calculators. Output is the whole point, so attendees iterate on what comes out rather than what gets stored.

A non-coder documented building the Wordle clone above end-to-end in 30 minutes with v0.dev plus shadcn — well under one workshop block [35]. The “boring random name” generator pattern shows up across Roblox Name Generator [28], My Poké Creator [30], and the wheels [20] [21] — all built on the same randomChoice(array) skeleton. Lovable explicitly publishes a “Build Interactive Games in Minutes” page [39] and a step-by-step gradient-generator how-to [43]; Bolt launched an official community Gallery in 2025 specifically to surface this scale of project [41]. For a more curated source pool, Questera’s 2025 ideas list adds AI meme generator, mood-playlist generator, interactive story engine, and generative art gallery in the same band [40].

Menu 3 — Fetch-and-render with one no-key API

Promote attendees from “stuff stored in my browser” to “I called a real internet thing”. Hard rule: one API, no auth, no key juggling. Mixed Analytics’s 2026 curated list is the canonical menu instructors pick from [67].

Open-Meteo is the strongest single starting point — explicitly no API key, no sign-up, no credit card with a CC-BY JSON feed and 10K free daily calls [44], and there is a March 2026 DEV tutorial that walks a layman through the entire fetch [45]. PokéAPI is the second-most-cited beginner target [46] with James Q Quick’s vanilla-JS Pokédex as the canonical short tutorial [47]. The full archetype set is canonised in Strapi’s 2026 portfolio list (weather dashboard, movie search, currency converter, recipe app) [68] and DEV’s “12 free and fun APIs” mapping [69].

Menu 4 — Scratch-an-itch apps (with real human stories)

The most motivating menu for laymen, because the build is about them. Use these as inspiration framing during the workshop’s “what should I build?” opening.

The pattern across these cases: a specific person solving a specific problem in their own life, not a generic tutorial example. Roose’s LunchBox Buddy shipped in roughly 10 minutes [83]. Patil’s baby-name ranker — ELO scoring, 10-accent pronunciation preview — took 75 minutes on Lovable’s free tier [75]. Peninsula SD reports building tools that replaced $30-40K commercial products in a few hours [78]. Microsoft’s case study has a non-technical employee shipping a 2-hour expense app and parents-with-kids building lemonade-stand managers and “nightmare-management” apps in ~20-minute sessions [81]. The Vibe Coder Blog’s 2026 kids guide names the four archetypes that work best for the youngest builders: flashcard quiz, collection tracker, tiny browser game, gratitude/mood journal [85].

Use these stories at the start of the workshop ⚠ before participants try to invent something. The single biggest scope-down lever is “build the thing you genuinely need this week” — it limits feature creep because the builder knows when it is enough.

Menu 5 — Micro-games (under 200 lines of state)

Games are the most popular ask in mixed-audience workshops, so have a short list ready. Stick to single-player or hot-seat; never real-time multiplayer [91].

The imagi × Lovable Hour of Code canonised the falling-objects catcher as the first vibe-coded game template — class-tested across schools — and the prompt that works is concretely “create a game where a cat catches falling treats and earns points” [36]. madewithlovable’s entertainment showcase enumerates Pong, Asteroids, CandyClicker, Hangin’ Man and Anxiety Balloon Pop, all hosted at public lovable.app URLs [25] [26] [27]. Cursor-built quiz games scaled to 7 hours for 100 AI-generated questions [37] — drop to ~20 hand-picked questions to land at 2-3 hours.

What doesn’t fit — and the swap

When an attendee proposes any of these, swap to the right-hand column on the spot. The same patterns repeat across every instructor write-up surveyed.

Sharetribe’s 60-hour marketplace post-mortem is the clearest case study: a forum has one user type, a marketplace has three — buyer, seller, admin — and the resulting build cascaded with seven consecutive “fixed!” declarations on one booking bug [87]. Roobykon adds the payment-edge-case angle: 45% of AI-generated code reportedly contains an OWASP Top 10 vulnerability, and AI consistently skips idempotency keys, webhook signatures, refund and dispute paths [88]. For mobile, the App Store memoir is unambiguous: “30% of the time prompting the AI for code that works and 70% of the time debugging various errors that show up due to the AI constantly losing context” [90]. Even Twilio’s own vibe-coding blog admits the setup (account, API keys, appsettings.json, terminal comfort) is itself a barrier for beginners [93].

Hackathon facilitators report scope is the single biggest demoraliser; the facilitator’s job is explicitly to narrow it, favouring visual / user-facing features over backend depth [3]. Jiang at 65labs frames the cultural rule: “building is so fast now that it’s often smarter to start over than to patch a messy product” — so accept the scope-down rather than fight it [4].

Picking one app for your workshop

If you have to commit to a default, pick from this short list. They are picked by ratio of “successful ships” to “stuck attendees” across the sources surveyed.

1. Tip / split calculator — for a mixed-audience adult workshop. Single screen, instant satisfaction, mobile-first feels real [13].
2. Habit tracker with emoji + 30-day grid — for participants who want something they will actually use. localStorage only [71].
3. Open-Meteo weather card for “my city” — for the one-API milestone. No key, deploys cleanly [44] [45].
4. Falling-objects catcher — for kids, classrooms, anyone who wants a game. Class-tested as Hour-of-AI canonical [36].
5. A scratch-an-itch tool the attendee names in the first 10 minutes — overrides everything above. Patil and Roose both picked things that mattered to them, and that motivation carried them through [75] [83].

Final rule of thumb for sizing → if the attendee cannot describe the idea as “one screen with [list of widgets] and a button that does X”, it is not workshop-scale; help them shrink it before any prompt is typed [7] [12].