Your assistant: the brief and Talk it through

HeroMind has one assistant, and it lives inside the app. It is not an agent you run elsewhere. It reads the same record your connected agents read, plus your tasks, goals and calendar, and it shows up in two places: a short brief at the top of the Planner each morning, and Talk it through in the sidebar under Daily Execution. Your agents feed it; it speaks.

Today's brief

A card at the top of the Planner tab, about a hundred words, written for the day you are looking at.

  • What matters. Up to three tasks, each with a reason in plain words: part of a quest, due this week, flagged by you. It starts from what you flagged. Carried tasks are never called missed; carrying a task forward is planning, not failure.
  • Real deadlines this week. Tasks with a due date inside seven days, at most two.
  • A thread you were working out. If you left an Ask conversation open in the Library, one line names it with a link to continue.
  • Your record. How many proposals are waiting on the Desk, how many the curator has grouped into themes, how many it set aside as chatter.
  • Capacity. Only when your calendar or time blocks are on the grid and your flagged tasks carry estimates: how much free time is left against what you planned.
  • What arrived. Tasks, sources and cards your agents filed since yesterday.
  • One question, at most. Reply to it in the card; the answer goes to Talk it through.

Every line has a dismiss with a scope: this item today, for the month, or forever. Refresh rewrites the brief from the current state; Hide removes today's; Turn off stops the brief until you turn it back on in Settings. The brief is also posted into Talk it through as the first message of the day, so the conversation always starts from it.

Talk it through

A conversation with the assistant about your day, your goals, or something you are reading. Ways I can help at the top lists what it is good at, grouped by Planner, Library, Decisions, and Goals and direction, with a For today row personalised from what is on your plate.

What it can do in the conversation:

  • Add a task when you ask for one, directly, with an undo. It never creates a task you did not ask for.
  • Capture a brain dump: anything you mention in passing goes to the Planner's review grid, where nothing applies until you confirm it.
  • Propose a next step toward a quest, naming the milestone and a Library page worth reading first. Add to my day accepts it.
  • Find my direction walks a short reflection and, if it lands on something, offers to Park it in Quests as an idea.
  • Answer from the Library, with citations, when the question is about something you have filed. Time questions work too: "summarize what I worked on today" lists every conversation and everything the librarian ran for you in that window (research, compiles, editorials), in your timezone, instead of picking the closest match.
  • Change its own settings in plain language: "keep briefs short", "no brief on weekends", "I schedule tasks for the day I do them".
  • Open a page for you: a quest, a Library piece, a settings section.

Personalize my assistant

Settings → Personalize my assistant. Every control is a real rule the assistant enforces, and the summary underneath says what it will do.

  • How I plan. I carry tasks forward and flag today's focus (the brief starts from flags) or I schedule tasks for the day I intend to work on them (the brief reads today's dated tasks as the plan). Neither treats carryover as a missed deadline.
  • How proactive to be. Help only when I ask (no brief) or A short brief each morning. Weekly review and accountability are not built yet and stay off.
  • Go into detail when it helps. Longer briefs when there is more to say.
  • Quiet hours. No brief is generated inside them.
  • Patterns from your history. The planning facts drawn from your own task history, with no model involved ("Personal tasks usually get done about two days after their date"). Advise me uses them; the assistant's brief does not. Dismiss anything you do not want kept.
  • Access and actions. What it reads, what it does directly, what it puts into review, and what it never does.

Rules worth knowing

  • It never comments on carried tasks and never cites a past due date as a reason to do something. If it does, that is a bug; tell it.
  • It reads your tasks, goals, routines, calendar, and the sources, pages and record entries visible to AI. It never reads the Notes tab, private items, or, in this version, your journal.
  • It writes to your record only as a proposal. Nothing it learns becomes current without your yes on the Desk.
  • Every model call it makes is logged under Settings → AI activity, with cost. HeroMind pays for these; you never need an API key.
  • Accountability, a weekly review, and delivery by email or push are not built. They are planned for after the assistant has earned a few mornings of trust.

Updated 2026-10-11 · All chapters · API reference