ChatGPT-refine →
Lovable-build
The chef's intro. For a 2–3 hour workshop with laymen: refine the idea into a structured PRD in ChatGPT (free, no Lovable credits burned [1]), then hand off to Lovable's Agent mode for the build. If every attendee already has a Lovable account, skip ChatGPT and use Lovable's built-in Chat Mode for the same conversation — it's purpose-built to precede a build [5]. The ChatGPT step earns its keep when attendees know ChatGPT but not Lovable, or you want one tool for ideation across the cohort before anyone signs in.
Mise · en · placeIngredients
-
1×
ChatGPT account free tier is enough — refine is free
-
1×
-
≈ 600cohort credits
~30/seat × 20 attendees -
4planning answers
what · who · why · key action -
2design buzzwords
from the spice rack below -
3+3+3real product names, testimonials, menu items
NO lorem ipsum — it cascades [3] -
1
-
1
≈ 2 credits per attendee.
That's 1 landing page + 4 edits.
Tell them. They'll spend better.
Two stages · one paste betweenMethod
Refine in ChatGPT ≈ 30 MIN · FREE
Goal: one structured paste for Lovable. The handoff is one-way — once you've seeded the project, Lovable's own docs say you must open Lovable to continue editing [1]. So the first paste matters more than any later one.
Have attendees answer Lovable's four-question planning frame [3]:
Then ask ChatGPT to expand into a PRD-style prompt — the meta-prompt is in The Master Sauce below. End with the magic tail: "Ask me any clarifying questions before you start" [3]. Single highest-yield trick in the docs.
Build in Lovable ≈ 60 MIN · ~2 CREDITS
Two routes from chat to a running app:
@Lovabletag inside ChatGPT — one auth grant; the conversation context becomes the initial build; no credits until build fires [1].- Copy/paste into Lovable's "new project" box — two windows, one paste; same cost [2].
Either way, the first paste lands in Agent mode (a.k.a. Build mode) — the execution mode that actually writes files [6].
Iterate by component SURGICAL · ~0.5 CR EACH
Build by component, not by page. The docs are emphatic: "build by component, not full pages at once", and "use real content instead of placeholder text" [3]. Asking for "the whole landing page redone" burns 2+ credits and tends to regress unrelated parts; asking for the hero CTA costs ~0.5 [10].
Drill the edit template — one sentence, two slots:
"Change [specific element] to [new spec]. Keep [what stays the same]."
Switch to Chat Mode for deeper changes
For structural rework mid-build, leave Agent mode and re-enter Chat Mode. Conversational, no code touched — "having a planning session with a technical co-founder before diving into development" [5].
The docs frame it cleanly: "Plan mode is for decision-making. Agent mode is for execution" [6].
A meta-prompt to start with
Don't want to write your own? Use the Lovable PRD Generator Custom GPT [15]
Workshop timing — the cooking schedule
Chef's notes
Fix-one-break-ten
The dominant Reddit complaint: "AI is still just soooooo stupid and it will fix one thing but destroy 10 other things in your code." [9]
Two-strike rule
Lovable's own guidance: three failed corrections = stop and reassess [12]. For laymen the rule is harder: two failed corrections → revert and re-prompt from scratch with more context.
Vague prompt → vague app
"A vague prompt leads to assumptions, and assumptions lead to apps that miss the mark" [5]. The whole reason the ChatGPT step exists.
Security blind spot
Independent assessment: 170 of 1,645 Lovable-built apps exposed personal data [9].
The lorem-ipsum trap
Placeholder text cascades — the AI builds layouts around the placeholders rather than the real content [3].
Label-rot ripples
Small label changes can ripple through file paths [8]. Name panels, buttons, cards once. Stick to it.
House
Rule
Plan the whole thing in chat.
Build one component at a time in code.
From the pantry
Cited sources · 17
No credits until build fires; further editing requires opening Lovable.
Four-question frame, edit template, clarifying-questions tail, real content rule.
Canonical long-form reference: meta-prompting, plan-then-build discipline.
Free: 5 daily credits, 30/month cap. Pro: $25/month for 100 credits.
Stable naming discipline; label changes ripple through file paths.
Per-prompt credit costs: ~0.5 styling, ~1.2 auth, ~2 landing page.
Concrete budgeting-app starter prompt: audience + CTA + aesthetic.
Three-strike correction rule: stop after three corrections with no improvement.
Real beginner workshop (Ljubljana / AI4DH, May 2026), no programming required.
Buzzwords interpreted as parameters: typography, spacing, shadow, radius, palette.
Beginner course repo — usable as workshop example bank.
Published course establishing the "first app in 1-2 hours" beginner norm.