One familiar moment helped shape the idea behind Dottie AI: you reach the end of a busy day, sit down, and struggle to remember what filled it.
There are a few completed tasks and some minutes on a timer. But there was also a plan that changed, a small problem you finally solved, and a moment when you thought, “I’m tired today.” None of those necessarily calls for a full journal entry. You might still want to remember them.
We want Dottie to make room for those fragments. You can leave a short note without organizing it first, return later to make sense of the day, and turn a useful thought into a plan for tomorrow.
Using that idea in everyday life reveals much more than a polished demo can. TestFlight feedback has helped us see the moments when the experience becomes confusing, awkward, or harder to trust than it should be.
These notes cover development through August 15, 2026. Dottie AI was in TestFlight beta during this period. The work described here was at different stages of development and testing; the notes for an installed build remain the reference for what it includes.
Make the conversation comfortable to use
For a chat window meant to capture everyday life, the reply matters. So do the few seconds after you press Send.
Did the message go through? Is Dottie answering, or waiting? If you scroll up to reread a note, will the screen suddenly pull you back to the bottom? When you reopen the app, can you find the conversation you were having?
You should not have to investigate those questions. Yet each one can affect whether you feel like recording the next small thing that happens.
From late July into mid-August, we worked on chat-history presentation, the way replies appear as they arrive, and scrolling behavior. Text should be comfortable to follow as it appears. Reading an earlier message also needs to feel deliberate, rather than like a contest with the newest reply. Those details become more noticeable as a conversation gets longer.
Input deserves the same care. Speaking may be easier when your hands are occupied; typing may suit a quiet room. We continued refining voice-input progress cues so that recording and reviewing the result do not depend on guessing what a button is doing.
These changes are difficult to package as one big feature. Their value is smaller and more immediate: one less moment of hesitation, and perhaps one more note you decide to keep.
Turn a request into a task you can understand
Conversation can be informal. A task or reminder needs more precision.
Suppose you say, “Remind me to look at the presentation tomorrow afternoon.” The intention is clear enough, but an exact time or duration may still be missing. If the assistant simply says “Done,” you have to leave the conversation and find out what that means. When was it scheduled? What kind of task was created? Is another step waiting for you?
Task confirmation has therefore remained an important part of our testing. You should be able to see what is proposed, check the relevant details, and understand whether an action is waiting for approval or has actually happened. The words in the conversation should match that state.
Our late-July development work included repairs to task-confirmation routing. Through early August, we continued refining chat actions and the transitions between their different states. The purpose is practical: you should not need to guess what the assistant did or repeatedly leave the conversation to make sure.
There should also be room to change your mind. You may want a different time, a smaller task, or no new task at all. Dottie can help organize an intention without treating every unfinished thought as a final decision.
A review needs to belong to the right day
Before a review can be insightful, it has to describe the right things.
If you selected yesterday, it should stay with yesterday. If a task is unfinished, it should not become a completed achievement in the report. If you recorded very little, there is no reason to manufacture a full day’s story around it. A gentle tone is welcome; accuracy is essential.
Since late July, our work has included grounding reviews in task information, correcting weekly metrics, and improving the handling of historical reports and date boundaries. By mid-August, finding past reviews and keeping the selected date aligned with the report were still receiving close attention.
These distinctions are especially important when you write a note late at night, return the next morning, or revisit an earlier day. “Which day does this belong to?” cannot always be answered by looking at today’s date.
Once that foundation is sound, a review can address more interesting questions. What progress was easy to overlook? Which task kept getting postponed? Is there one specific thing worth trying tomorrow?
We want the daily review to connect its suggestions to the day you actually recorded. A small, recognizable next step is more useful than another general list of ways to become a disciplined person.
Keep improving the rest of the app, too
People do not use the AI and the ordinary app as separate products. You remember something in a conversation, check the task on the home screen, start a timer, and look at its progress on the Lock Screen. Those actions are part of one day.
Work during this period therefore extended beyond chat. Early-August changes included home-screen and statistics interactions, along with timer behavior around screen locking, syncing, and moving between views. Mid-August customer-feedback work also addressed task controls, tags, and historical reviews.
For a timer, the questions are direct: is it still running? Why does the Lock Screen show a different time? Did syncing affect this session? The internal explanation may be complicated, but the experience should give you a clear answer.
New features can make an app worth trying. Reliable everyday behavior helps make it worth keeping. We need to pay attention to both.
How a report becomes an improvement
Some feedback is easy to describe: a button does not respond. Other feedback begins as a feeling: a sentence sounds wrong, a review misses the point, or an action leaves you unsure what happened.
Both are useful. One helps us locate a fault. The other exposes a gap between how we understand the product and how someone experiences it.
Here is an example of a helpful report:
I wanted to schedule reading for tomorrow. After Dottie replied, I thought the task had been created, but I could not find it on the home screen. I’m not sure whether I missed another step.
This is an illustration, not a quotation from a tester. It connects the intention, the interpretation, and the result. That gives us a path to follow: was the explanation unclear, did the action fail, or was the result hard to find?
A build number, approximate time, and screenshot with private details removed can help. You do not need to record your entire day or identify the technical cause before contacting us.
From there, we try to reproduce the situation, make a change, check it, and observe it again in testing. Some issues have a direct cause; others need more work across devices and conditions. Changing the code does not mean we should stop paying attention to the original scenario.
Keep testing with an ordinary day
Dottie AI is still taking shape. Natural conversation, useful suggestions, and recognizable reviews need to be tested against real routines, not just ideal examples.
If you would like to participate, the beta guide explains the current way to join. Begin with a few things you would otherwise forget: an email finally handled, an interrupted reading session, or a plan for the next morning. When you look back, ask whether Dottie helped make the day clearer.
You do not need to become a professional tester. Use the app normally and tell us what confused you, what you expected, and what felt cumbersome. We will keep documenting the changes so that each update answers a simple question: what, specifically, became better this time?