RexiO closed test goes live on 10 July 2026 — 500 users, real conversations, real signal

We're opening RexiO to 500 people on 10 July 2026.

Not a teaser. Not a waitlist. A real closed test, with real users, on the product we've been building for the last 18 months. And yes — 500 is small, deliberately. We've spent the last six months scaling from a private alpha of 30 users to where we are today, and we know exactly how much our system can hold without breaking.

What this test is

The closed test is the first time RexiO will run at anything approaching real production load. The 500 people in this group are split across three groups:

• ~250 returning alpha users — people who have been with us since the early days and have given us months of feedback. They know what RexiO does, and more importantly, what it doesn't.
• ~200 new users from Bangladesh — most of them on mid-range Android, on 3G/4G, with Bangla/Banglish as their default input. This is the demographic we built for.
• ~50 users from outside Bangladesh — for cross-language and cross-region testing. They help us catch the edge cases that a Bangladesh-only test would miss.

What we're testing

Three things, in order of priority:

1. Conversation quality at scale. Does RexiO stay coherent across long sessions? Does it remember what you said 20 messages ago, not just what you said in the last message? Does it actually know you?
2. Bangla and Banglish in the wild. Real users don't type like our test prompts. They mix scripts, swap between Banglish and English mid-sentence, and ask questions our QA never thought of. We want to break our Bangla tokenizer, then fix it.
3. The boring stuff. Memory. Token budgets. SSE stream stability. Mobile battery. What happens when 500 people all hit the same promo link at once.

We're not testing new features. We're testing whether the version we have can carry the weight.

What this is not

We're not opening sign-ups. If you're not in the test, you can't join. We know this is frustrating, and we'll be expanding the test over the next two months as we fix what breaks. The waitlist exists, and it works — but we won't be opening it again before launch.

We're also not testing a "lite" version of RexiO. The 500 people in the test get the full product. That's the only honest way to know if it works.

The next 60 days

If everything holds:

• 10 July 2026 — closed test opens to the 500-user cohort.
• End of July — first round of fixes from the test, retest with the same cohort.
• Mid August — expand to 2,000 users, opened from the waitlist.
• End of August / early September — public launch, if the test data supports it.

The public launch is contingent on the test going well. We'd rather push it back a month than ship something that doesn't work for Bangladesh the way we want it to.

Why 500

We've seen products scale to 50,000 users on day one and then spend six months cleaning up the mess. We'd rather run a real test on 500 people, get the data we need, and then scale with confidence. The number isn't arbitrary — it's the largest cohort we can support with our current infrastructure while still having headroom to absorb the failure modes we expect to see.

What to expect if you're in the test

• You'll get an email on 9 July with a sign-in link and a one-page guide to what RexiO does and doesn't do.
• We'll ask for feedback in-app, in Bangla or English, your choice. Every piece of feedback gets read.
• If something breaks, we'll tell you. We're not going to pretend the test is going perfectly when it isn't.

If you're not in the test

There's not much to do right now besides wait. The waitlist is open at rexio.pro, and we work through it in order. The next expansion is mid-August. If you want to help us in the meantime, the best thing you can do is tell us what you want RexiO to be — the form on the waitlist page is short, and we read every submission.

— The RexiO team