The starter template for AI startups. Auth, checkout and i18n included. Français
← All posts Strategy

Why your AI startup shouldn't start from a blank repo

Alinement team Alinement team September 8, 2026
ai-startupsshipping

Every AI startup has two products. The first is the thing you pitch: the model, the workflow, the insight nobody else has. The second is everything around it: the marketing site, sign-in, checkout, analytics, translations, SEO, a deploy pipeline, and the tests that stop all of it from breaking at 2 a.m.

Nobody buys the second product. But nobody can buy the first one without it.

The month that disappears

Start from git init and the first few weeks go roughly like this:

  • Week 1: a landing page, a pricing section, a logo that looks fine in dark mode.
  • Week 2: sign-in. Passwordless is nicer, so now you need email codes, resend timers, expiry, and error states for the bad paths.
  • Week 3: checkout. Verify the email before taking money, add coupons for your launch, wire up a payment provider, and send ecommerce events so you know where people drop off.
  • Week 4: translations for your second market, Open Graph tags, structured data, a sitemap, and a test suite you'll run before every release.

That's a month before you've written one prompt. Your competitors aren't waiting.

What starting on Alinement changes

Alinement is that whole second product, already built and already tested. When you clone it you get:

  • Marketing pages, a blog and guides with English and French built in, and a check that fails when a string is missing in either language.
  • Passwordless auth pages with 6-digit codes, a 60-second resend countdown, and simulated error states you can trigger in demos with addresses like you+no-network@example.com.
  • A gated checkout: the email is verified before payment, coupons apply before the total, and card, PayPal and Flutterwave inline checkout are wired in. GA4 tracks add_to_cart, begin_checkout and purchase with stable item IDs.
  • SEO handled: canonical URLs, hreflang alternates, Open Graph cards, product and article structured data, and a sitemap that skips pages marked noindex.
  • Quality gates that run before each commit: Biome, Knip, Playwright, a link checker, a coverage floor, and Lighthouse with a score of 90 or higher.

Built for the AI part

The template doesn't pick a model for you. All client calls go through one file, src/js/dataservice.js, backed by a small API client that handles timeouts, cancellation and per-request reCAPTCHA tokens. In demo mode it returns realistic fake data. Set DATA_MODE=live and API_BASE_URL, and the same calls reach your backend, which can sit in front of whichever model provider you choose.

That makes the boundary simple: the UI never needs to know whether a response came from a fixture, your API, or a model.

Spend the month on the moat

The plumbing matters, but it doesn't set you apart. Start on Alinement, deploy on day one, and put the month you'd have spent on checkout edge cases into the product only you can build.

If you'd like the services connected for you as well, the Business owner plan covers that setup.

Share this
Launch offer: 20% off either licence with code LAUNCH20 at checkout.