Lovable or custom development: the real trade-off

The question isn’t which one is best, but at what moment. A generator proves an idea in a few days; collecting money from real customers takes something else. Four paths, the same six criteria, ours included.

The six criteria

• Specialization: Does this model redo products every day, or once in a while?

• Fixed price or time and materials: A price written down in advance, or a meter that keeps running as long as it lasts?

• Team size: How many people, and what happens the day one of them leaves?

• Entry price: What you need to put on the table to get started, excluding subscriptions and hosting.

• Code ownership: Who owns the code, the access and the data, and from when?

• Maintenance: Who answers, how fast and at what price, six months after delivery?

What each one does, and doesn’t do

The AI generator alone

You describe the app, the tool writes it, and it runs online the same day. It’s the shortest path between an idea and a working screen.

• Specialization: The prototype: screens and flows

• Fixed price or time and materials: Subscription, then pay-as-you-go credits

• Team size: You, alone, in front of the tool

• Entry price: Depends on the credits used

• Code ownership: Exportable code, accounts with the vendor

• Maintenance: Every fix consumes credits

When it’s the right choice

• Checking that an idea interests someone, before investing.

• Showing a complete flow to a client or an investor.

• An internal tool for five people who know each other.

What this model doesn’t do

• Without automated tests, one fix brings back two more.

• The prototype holds up a demo, not a load.

• “My €50 a month turned into €300 to €400 of credits.” A founder’s own words, quoted as is.

Check before you sign

• The data access rules (RLS) have never been checked.

• API keys are readable from the browser.

• You’re already taking payments on a prototype.

The generator plus an hourly developer

You keep generating, someone catches what breaks. It’s the most common setup right after the first real users.

• Specialization: Depends on the person

• Fixed price or time and materials: By the hour or by the day

• Team size: One person, on and off

• Entry price: Depends on the profile

• Code ownership: In the tool until the exit is done

• Maintenance: On demand, no commitment

When it’s the right choice

• Unblocking one specific point: a payment, an integration, an access rule.

• Holding out until a deadline: a demo, a funding round, a trade show.

• Keeping control of the screens while someone secures the rest.

What this model doesn’t do

• Nobody takes over the whole thing: they repair where it breaks.

• The tool can regenerate over what was just fixed.

Check before you sign

• The fixes aren’t versioned anywhere outside the tool.

• No tests are written, so nothing protects the fixes.

• The same outage comes back a second time.

Custom from day one

The product is written in code from the first day, without going through a generated prototype. Slower to see on screen, sturdier to carry.

• Specialization: The product, from design to production

• Fixed price or time and materials: Fixed price or time and materials, depending on the provider

• Team size: From one to several people

• Entry price: Varies by provider; with us, an online estimate then a fixed price in the quote

• Code ownership: Yours, if the contract says so

• Maintenance: Separate contract, to plan from the quote

When it’s the right choice

• The idea is already validated: customers are paying, or the waiting list is full.

• The product touches money, health or personal data.

• A large account asks you where its data is hosted.

What this model doesn’t do

• Slower to see on screen than a prototype generated in an evening.

• More expensive to enter: not the right way to test an idea.

Check before you sign

• You have no users yet: validate before you build.

• The quote includes neither data migration nor documentation.

• Nobody commits to what happens after delivery.

The managed exit (us)

We keep what the prototype proved, get it ready to take money from real customers, then move it to infrastructure that belongs to you.

• Specialization: SaaS takeover and redesign. Nothing else

• Fixed price or time and materials: Fixed price. Audit billed separately

• Team size: A senior founder, surrounded by trusted specialists

• Entry price: Online estimate, fixed price in the quote

• Code ownership: Next.js and Supabase, on your accounts

• Maintenance: 3 to 20 days a month, as a flat plan

When it’s the right choice

• The prototype works and the first paying customers are arriving.

• You want to keep evolving the simple screens yourself.

• You don’t know whether your app is really at risk.

Our limits, said out loud

• A small senior team. Not ten, not thirty.

• No showcase sites, no e-commerce, no games.

• The audit is billed separately: it isn’t deducted from the fixed price.

• It isn’t a penetration test, it’s a health check of a product.

When it isn’t us

• Your prototype has no users yet: validate first, move out afterwards.

• You’re looking for the cheapest solution right now: that isn’t us.

How to combine them

• Generate first, code afterwards: The two paths aren’t opposed, they follow each other. The generator is for finding out what to build; the code is for building it for real. The tipping point is the first paying customers.

• Keep the tool for the simple screens: After the exit, you keep evolving the content and the simple screens yourself. The rest goes through us, or through any developer who takes over the code.

• The audit before the decision: Twelve checkpoints (access rules, exposed keys, personal data, payments, performance) tell you in five days whether your prototype can open to real customers, or not yet.

• Validate before you build: If nobody has paid yet, neither the generator nor custom development is the right spend: a page, a waiting list and an honest measurement cost less than a product nobody is waiting for.

Where the numbers come from

• 45% of AI-generated code fails security tests (Veracode, 2025).

• Vulnerability CVE-2025-48757 concerned the data access rules of generated apps.

• Our simulator gives a range after a few questions; the fixed price is written in the quote.

• We don’t publish any price range we haven’t collected ourselves: it would be more convenient, and dishonest.

• No company is named in this comparison. This isn’t a legal precaution: a brand name says nothing about what happens once the quote is signed, a way of working does.