On July 9 I posted a link on my own accounts and asked people I know to come and break the thing I had been building. That was the day ParrotTalk stopped being a private project. It has been an open, public beta ever since, and the doors have simply never closed.
It is a strange feeling to recommend your own unfinished work. Up to that point every bug had been mine, found by me, at my own pace, in my own browser. Handing it to real people means handing over the pace too, and finding out which of your confident assumptions were merely convenient.
Removing the step where people give up
The first thing the beta taught me had nothing to do with the AI I had fussed over. It was about the path to the product. My tidy plan of emailing each tester an invitation and a test sheet had a dozen places to die: a promotions tab, a phone read meaning to open it later on a laptop, a friend-of-a-friend with no email at all and no way in.
So I put the whole tester kit straight on the home page, no email required and no permission needed. Anyone who reached the site could start. That is a pattern I have kept ever since, and it is the same instinct that keeps the site 100% free with no sign-up: I would rather have people practising than have their contact details.
What real people actually did
The reports were nothing like my own testing. Nobody followed the happy path I had walked a hundred times. People clicked in orders I had not imagined, on browsers I had not opened in weeks, and stopped where I had assumed the next step was obvious. Almost none of it was about the AI. It was about the plumbing: where do I click now, did that save, is it still loading.
That is the real value of opening early, and it is worth more than any bug list. A few days in, I had gone comprehensively blind to my own product. Every screen was obvious to me because I had built it, and obvious-to-the-builder is the least useful kind of obvious there is. It takes about ten minutes with someone else's confusion to get your eyes back.
Where this is going
Being a public beta is not a stage I am rushing to leave. It is an honest label: the tool works, it is genuinely useful, and it is still improving in the open, with real people shaping it. Calling it finished would be the dishonest move.
There is one next step I am planning, and I want to describe it without overselling it, because the whole project runs on not overselling. I would like to invite volunteers who sit both a full ParrotTalk mock test and the real exam to share their official result, so I can measure how close the estimate really gets and publish that openly. It is fair to make it worth their time, and I intend to. I am not putting a date or a number on it yet, because I will not announce either until I can stand behind it.
The doors are still open. No account, no wallet, no waiting list. It is a public beta, and you are welcome in it.
Opening up early was the best decision of the project. It turned a thing I was sure about into a thing other people could actually use, which are not remotely the same thing.
The quickest way in is still the free Quick Mock Test, all four sections in about 50 minutes.