@RhysSullivan Agree, more to teach. A tiny recipe in the docs would do it: "memory in 20 lines with the primitives you already have". Most people copy a working example long before they compose from primitives.
@yz_aisdr 16 signups, 0 paid is a normal week one. One thing worth ruling out before blaming intent: go through your own checkout with a fresh email in a private window, start to paid. If it stalls anywhere, it's not the curiosity.
@Denche2005 Because the 10% Claude doesn't write is knowing what breaks when a stranger uses it. That's what "experience" means on a hiring call. Ship something small, watch a real user get stuck, fix it. That story beats any degree.
The fix: give the new column a default or let it be empty, or update the trigger so it fills it. Then sign up once with a brand new email and check that the profile row is really there.
Signup fails with "Database error saving new user". Your own account works fine.
Often it's the trigger that creates a profile row. You added a required column later, the trigger never fills it, so every new signup breaks.
Tried a fresh email since your last schema change?
@uriel_builds@X Building FirstUser: an AI test user that signs up, logs in and tries checkout on your Lovable, Bolt or Cursor app, then hands back a fix prompt. https://t.co/AOYFdwRCYv
@systemdesignone Fastest review prompt I use: "List every place a logged out or brand new user could hit this code path, and what they would see." It catches the bugs a normal review skips, because it reads the diff from the stranger's side.
The fix: set success_url to the exact same host as the login cookie, www or no www, not a mix. Then complete one real payment with a new account and check that you arrive already logged in, on the paid plan.
They paid. Then your app treated them like a stranger.
Stripe sent them to www, but their login cookie sits on the bare domain. Or vice versa.
Payment went through. The session didn't. They land logged out.
Have you finished checkout as a new user, not already logged in?
Before you decide nobody wants your app, check that a stranger can get in. Fresh email, live URL, private window. A confirmation link pointing to localhost or a profile trigger throwing "Database error saving new user" looks exactly like "no signups".
@ArpokratPrivacy The flip side shows up in the same apps: once RLS gets added to close the leak, the signup or profile insert often starts failing for real users. Worth re-testing signup and login as a brand-new user after every security fix.
@Jonatech256 One hour solo and it renders live, nice. If you keep building it, the first thing to check before sharing: can a stranger with a new email get from the live URL to their first proposal? Happy to run a free FirstUser test on it.
@codewithharshad Nice, live preview + copy code is the right loop. Preset idea: a subtle shake for a failed form input, like a wrong password. If Animably gets accounts or a paid tier later, happy to run a free first-user test of signup and checkout.
@arhamsai Respect for shipping at 17. If your SaaS has a signup or checkout, I'm happy to run a free FirstUser test on the live URL: it goes through it like a brand-new visitor and sends a screenshot of any step that breaks. No strings.
@LiveLoveSEO Agree for app #1. The moment you share the link, though, it breaks for a stranger who won't tell you. That's when a fresh-browser run of signup, login and checkout earns its 5 minutes.
@suraj_sharma14 Solid stack. One step I'd add before launch: open the live URL in a private window and go signup โ confirm email โ login โ checkout as a stranger. With Lovable + Supabase + Clerk + Stripe, the seams between tools are where new users get stuck.
@CodeWithGD True. And there's a quieter version of that trap: some people do want it, click the link, and can't get in. Confirmation email never arrives, login loops, checkout never upgrades the plan. Worth ruling that out before deciding nobody wants it.
@Denche2005 Moving countries at 12 without speaking the language is a harder cold start than any launch. Football to five languages to data engineer is a better origin story than most founder bios.
@TheJordanRules_ That's the right order. The part I'd still test is the later handoff: when that anonymous user finally starts a trial or creates an account, does their work stay attached? That's where a lot of first paid sessions quietly reset.
@systemdesignone One trick: I make the agent write the test plan before the code. Three lines: what a new user does, what they should see, what would count as broken. Then it builds against that, not against my prompt. Way fewer surprises after deploy.