June focused on the path from saying something to understanding what ISEA is doing with it. We refined voice recording, made Agent thinking visible across tool loops, measured the stages between a request and its response, and added iOS widgets for quick access to daily context.
The common thread was feedback. When an interaction takes a moment, the app should make that moment easier to understand instead of leaving you to guess whether it is listening, thinking, or waiting.
Voice capture feels more intentional
The Agent composer received a deeper pass on voice recording. Press-and-hold gestures, waveform feedback, cancellation, release behavior, and transcript handling were refined so recording has a clearer beginning and end.
The goal is not to make voice feel like a separate mode. It should be another natural way to start the same capture flow. You can speak while walking, describe what just happened, or leave a longer thought without typing it out first.
We also tightened the edge cases around gestures. Cancelling a recording should cancel it. Releasing should not lose a transcript that is already ready. These small details determine whether voice feels dependable enough to use in the middle of a real day.
Agent thinking is easier to read
Agent work can involve several tool steps before a final answer appears. In June, the client began showing thinking feedback across those gaps instead of only showing a loader at the start of a request.
Reasoning can now sit in its own expandable block, with a clearer distinction between active thinking, streamed text, and completed output. Markdown handling and the thinking indicator were also refined so the conversation stays legible while the Agent moves through a task.
This is a modest interface change with a meaningful product effect. When a request takes a little while, the user should be able to tell whether ISEA is still working, waiting for a tool, or ready with the result. Better timing instrumentation behind the scenes helps us measure those stages and improve them over time.
A glanceable layer on iOS
June also introduced a first set of iOS widgets. The widgets bring a small amount of ISEA context outside the full app: today’s task count, water progress, mood, habit streak, quick record actions, and a short companion echo.
They are designed for glanceable use rather than a second dashboard. A widget can point to the right record flow, open a quick action, or remind you what has already been captured today. The snapshot data is kept in sync with the home summary so the surface remains useful between app launches.
This is the beginning of a broader idea: ISEA should meet you in the places where a day already happens, while keeping the full record and review experience inside the app.
Seeing the path from request to result
Behind the scenes, client and server timing now distinguish the first reasoning signal from the first text token. That gives us a clearer view of where an Agent request spends time and makes it easier to improve the stages users actually feel.
We also reduced unnecessary waiting around message persistence and made the Agent runtime more flexible about thinking-provider options. The goal is not to expose more infrastructure. It is to make the visible experience more predictable when a request includes several steps.
More control around accounts and platforms
Alongside the visible product work, June tightened several foundations. Account cancellation controls became configurable, iOS permission prompts were localized, and the app moved to Expo SDK 56. The Agent runtime also gained more flexible thinking-provider options and a cleaner way to activate or deactivate skills.
These changes are not meant to add ceremony. They give the product more room to behave correctly across platforms, accounts, and different types of Agent work.
June was about making the Agent’s behavior more transparent and less intimidating: speak naturally, see when it is thinking, understand when a result is ready, and reach the right record action from the places you already look. The next step is to keep that clarity as the product handles more context and more kinds of daily work.