I built an AI regex tool for overseas users – here’s the stack and what’s working so far

Fair question — I don’t have a clean “who hits 8/day” pattern yet (still early). The working hypothesis is heavy iterators (tighten/loosen loops) and people regenerating near-duplicates. Going to watch that before changing the free cap again.

Credits pack is on the radar for occasional users — I’m holding it until I see how many people bounce at the monthly ask vs just needing a saved library. Glad the ChatGPT↔regex101 loop landed; that’s still the core wedge.

Really useful split — “hit the daily cap mid-task” vs “want a saved library” are different buyers. Right now I’m packaging Pro around library/history (+ unlimited as the relief valve) rather than inventing two SKUs before I see which motive shows up more.

This is the sharpest take in the thread. Generation isn’t the dealbreaker — confidence is. I’m planning must-match / must-not example lists (plus we already surface a light ReDoS hint) so people can prove a pattern before they trust it. Appreciate you framing it that way.

History/library is exactly the Pro bet now — local recent on free, cloud sync across devices on Pro. Thanks for naming it so clearly; that feedback moved it to the top of the upgrade copy.

You’re right — guest vs signed-in free vs Pro weren’t explicit enough, and the post still said 5 while the app was 8. I’ve aligned the live copy to the three-state line you suggested. Agree $9 is less the issue than whether repeat users see Pro value (esp. cloud history) before they hit the wall.

Quick update from the thread feedback:

The live free tier is 2 guest tries, then 8/day after email sign-in (the post’s “5/day” was stale — thanks for catching it). Pricing + the counter now say it in one line:

“2 guest tries · 8/day with a free account · Pro: unlimited + cloud history & share links”

Pro pitch I’m leaning into: cloud history / saved library first; unlimited is the safety net while iterating. Next build: must-match / must-not assertion tests so confidence doesn’t depend on a single sample string.

The dealbreaker for me isn’t generation, it’s confidence. I’d pay sooner for an adversarial test set: examples that must match, examples that must not, plus a ReDoS warning. A regex that looks right but fails on edge cases is exactly why people still bounce between ChatGPT and regex101.

The Paddle MoR note is gold for overseas indies — that’s a distribution problem people under-document. On pricing, I’d watch whether Pro buyers are people who hit the daily cap mid-task, or people who want saved libraries/history. Those two motives want different plans.

Solving that "ChatGPT + regex101" context switching loop is super slick. For $9/mo, it might be a bit steep for devs who only write complex regex twice a month. Have you considered a cheap one-time credits pack alongside the monthly sub? Might convert way more casual users.

I tried the guest flow, and the “2 trials left” counter makes the limit easy to understand. One thing confused me, though: the post says Free includes 5 generations a day, while the live pricing page says 8 a day after sign-in. That makes the $9 question harder to judge because people aren’t evaluating the same free plan. I’d make the three states explicit in one line near the counter: “2 guest tries • 8/day with a free account • Pro: unlimited, cloud history, and share links.” Then repeat it after a successful generation, when upgrading becomes relevant. $9 doesn’t feel obviously high to me; the bigger issue is whether repeat users understand the Pro value before they hit the limit.

The pricing question may depend less on whether $9 feels reasonable and more on what actually pushes someone past the free workflow.

Have you seen any pattern yet in what frequent users are doing that makes 5 generations/day meaningfully limiting, rather than enough for most regex work?

Looks nice!
My #1 feature would be a history or library I generated.
Many things I need multiple times, but with time between and building a library in the tool would be important to come back to the tool and continue using it