I wanted a clean landing page live the same afternoon I sketched it. Not a full product. Not a WordPress theme chase. A single public page with a clear headline, a short story, and a soft path into the rest of mustafa.net. So I one-shot it with Lovable, published the first pass, then spent the real time on what the generator never owns: auth edges, SEO basics, hosting, and branding that matches the site I already run.
This is the Builders version of that story — what shipped fast, what I kept, and when I reach for Lovable versus Cursor for mustafa.net-style work.
What Lovable actually shipped in one pass
Lovable is good at the first visible surface. I described the page in plain language: audience (homelab / builders), tone (concrete, no hype), sections (hero, what it is, what it is not, soft CTA), and a neutral color direction. The first generate gave me:
- A usable layout with hero, short body blocks, and a footer
- Responsive spacing that did not collapse on a phone
- Placeholder copy that was close enough to edit instead of rewrite from zero
- A deploy path I could click without wiring a repo by hand that night
That is the honest win. The Lovable one-shot site got me past the blank canvas. I was not fighting CSS grid for two hours before the idea existed in public.
What it did not give me out of the box: my existing brand tokens, my real nav, correct meta tags for how I publish, or any of the auth / membership edges I sometimes bolt on later. Those stayed on me.
What I fixed by hand before it felt like “mine”
Branding and visual continuity
The generated palette was fine for a demo and wrong for mustafa.net. I swapped fonts, tightened the hero, and aligned buttons and link color with the rest of the site. If a page does not look like it belongs next to Homelab Gear or the Local LLMs hub, readers bounce even when the words are good.
SEO and publish hygiene
I wrote the title, meta description, slug, and Open Graph bits myself. Lovable’s defaults were generic. For a Builders post or landing page that should rank on a money phrase, I treat meta as part of the ship checklist, not an afterthought.
Hosting and the real URL
The quick preview host is useful for feedback. Production still meant pointing the page at my real domain path, checking HTTPS, and making sure internal links hit live mustafa.net URLs — not preview URLs that die next week.
Auth and anything stateful
I did not need full auth for this page. When I have needed gated bits elsewhere, Lovable’s first pass is not where I finish that work. Cookies, callbacks, and “who is logged in” still get edited in a real codebase with Cursor (or by hand). Treat one-shot UI as the shell; treat auth as owned code.
When Lovable is the right tool vs Cursor
I keep a simple rule:
- Lovable when I need a public surface fast — landing page, concept page, marketing shell — and I am willing to re-skin and re-host afterward.
- Cursor when I am inside an existing repo: WordPress-adjacent tooling, interactive checkers, Compose files, Ansible, or anything that must match patterns already on disk. That is why the Cursor tool page sits in my tools shelf, and why the free local LLM hardware checker was a Cursor-style ship on top of the site, not a Lovable export.
For mustafa.net Builders work, Lovable shortens the path from “I should ship this page” to “there is a URL.” Cursor shortens the path from “this system already exists” to “the change is correct and reviewable.” I use both. I do not pretend one replaces the other.
What I kept vs what I threw away
Kept:
- Structure of the page (section order)
- The core promise in the hero
- The soft CTA pattern — link real tools and hubs, no fake urgency
Threw away or heavily rewrote:
- Generic stock phrasing
- Default colors and type
- Any generated “feature grid” that sounded like SaaS hype
- Preview-host links that were never meant for long-term SEO
Soft next step
If you are building in the same lane — local AI, homelab, honest tools — start from the Local LLMs hub or the hardware side on Homelab Gear. Use Lovable when you need the first public draft tonight. Use Cursor when the work lives inside the stack you already maintain.
FAQ
Can Lovable replace a real site build?
For a single landing page or concept shell, often yes for the first publish. For mustafa.net-style work with shared branding, SEO, and deeper tools, I still finish by hand or in Cursor.
What broke first after the one-shot?
Branding continuity and meta tags. The layout held; the “this is clearly my site” layer did not.
When should I skip Lovable and open Cursor?
When you are editing an existing codebase, shipping an interactive tool, or anything that needs auth, precise routing, or files that already live in git.