Meetings

A meeting is the thing you prepare for. Everything else hangs off it: the job advert or the prospect notes are scoped to it, and every call you run against it is recorded under it.

You can have as many as you like at once, and they can be of two kinds — an interview or a sales call. The kind decides what material the app asks for and what it calls things on screen.

From a terminal

Everything in the window works headless (that is a rule here, not a nicety), and meetings were the one thing that did not:

whispr meeting list                                   # and which one is selected
whispr meeting create --company Northwind --role "Staff Engineer"
whispr meeting create --type general --title "Team sync" --no-select
whispr meeting select <id>

create selects the new meeting unless you pass --no-select, and names it the way the window would. select takes an id — the same identity the app uses — because matching on a name would quietly choose a different meeting.

There is deliberately no --type on any change: the kind is fixed once material is loaded, for the reason above. Until this existed, a headless install could not select a meeting at all, so anything scoped to one — including whispr prep add --tailored — had nothing to attach to.

A second round is not a second meeting

A meeting has many sessions. When you interview again at the same company, start another session against the same card: the prep you already loaded is still there, and the history piles up in one place. There is deliberately nothing here that offers to duplicate a meeting for a second round.

What a card tells you

The title, the type badge, and whether it is the one currently selected. Under that, the role and company, when it is scheduled, and the website if you set one.

The counts line — materials and recordings — is a button. It opens History filtered to this meeting.

Beside it, when there is anything, "2 things still to do". That counts across every round of this meeting, not just the last one, and it is the same button: what is owed and where to go and do something about it are one click apart. When nothing is outstanding the line is absent rather than reading "0 to do" — the whole point is that the one card with something owed stands out from the ones that do not.

Last time shows the headline of the most recent debrief on this meeting. A second round shares this card and reuses its prep, so the minute before the next call is when the last round's conclusion is worth most — and it is exactly the minute it used to be nowhere on screen. The whole debrief is one click away in History.

Your own notes about the meeting show on the card too, so the field is worth using: who referred you, what the recruiter said, what was promised on the last call.

The readiness pills

Small labels saying what this meeting still needs. They cover material only — whether your keys work is a global question with its own screen.

There is one asymmetry worth knowing. The thing that follows you — your CV, or what you sell — only appears when it is missing. "CV loaded" on all eight cards is noise; "no CV yet" on all eight is one instruction, in the place you will see it.

A further pill appears once you have asked for an alignment, saying whether it is ready, still generating, failed, or out of date because the material changed under it. No pill at all means nobody has asked for one yet.

Creating one

Create meeting — in the title bar, and at the top of this tab — opens the wizard on a blank meeting. It heads itself Create a meeting, and the meeting you were on keeps everything loaded for it; nothing is renamed or rescoped by creating another.

The same wizard also opens on the meeting you already have, from the end of the guided setup or when you re-enter the flow. It heads itself Set up this meeting then, and Continue updates that meeting rather than making a new one. The heading is the difference, and it follows the radio buttons on that screen if you change your mind there.

Selecting, starting, editing

Select makes this the meeting that prep and the next session belong to, and takes you to the Prep tab. Open is the same button when it is already selected.

Start begins a session against this meeting directly, without changing tabs.

Both are unavailable while a session is running — switching meeting mid-call would rescope the thing being recorded. Stop first.

Edit changes the name, company, role, website, scheduled time and notes in place. It does not offer to change the type: material added under one kind would end up on the wrong side of a meeting that had become the other, and there is no way back from that.

Archiving

Archiving takes a meeting out of the list without removing anything. Its prep and its recordings stay exactly where they are, and Unarchive brings the whole thing back.

An archived meeting cannot be started, which is the point — it stops last month's card being the one selected at 09:59. When you have archived anything, a checkbox appears to show them again.

What this meeting is for

Every meeting card carries a short list of goals, and every engine is told them before it answers. An answer that can naturally move one forward will — and nothing is forced, because an answer that chases a goal past the point the conversation allows is worse than one that simply answers well.

There are two kinds, and the difference decides who can tick them off.

To cover — something to learn or say. "Learn their decision process." "Mention the migration we did for Acme." Whispr can hear these being covered and ticks them off for you, and says so on the card: a goal marked heard on the call was settled by Whispr, not by you.

To leave with — something the meeting has to end with. "Book the technical demo." "Agree a pilot start date." Only you can mark these. A transcript cannot know whether a commitment was real — somebody saying "yes, let's get that booked" is not a booking — and a meeting that ends believing it succeeded is worse than one that knows it did not.

Goals belong to the meeting, not to the call. A second round against the same company inherits them, along with the prep, which is the whole reason a meeting and a session are different things.

A completed goal stays on the card, struck through, rather than disappearing. Afterwards it is the most useful thing there: what this meeting achieved, and what it never got to.

They are optional. A meeting with no goals behaves exactly as it always did.

Starting from a template

While a meeting has no goals yet, the editor offers a few starting sets — discovery, demo, pitch, and one for an interview or a working meeting. One click fills the list.

They are starting points, not checklists to obey: every goal lands as an ordinary one you can edit or delete immediately. They are short on purpose, and they are written as things to find out rather than as jargon to recite — "what it costs them to leave it as it is" invites a conversation, where "qualify budget" invites an interrogation.

The offer disappears once you have goals of your own, and applying a template adds to what is there rather than replacing it. If you had already typed two goals you wanted both, and quietly discarding them would be the worse surprise.

Three kinds of meeting

Interview — you are being interviewed. Whispr grounds answers in your CV and the advert, and the checklist asks for both.

Sales call — you are selling. Answers are grounded in what you sell and what you know about the prospect, and an invented claim is treated as the commitment it would be.

Meeting — everything else. A team sync, a client check-in, a conversation you want recorded and helped with. It asks for no material at all, because a team sync needs neither a CV nor a product sheet, and a checklist demanding one would tell you a perfectly prepared meeting was incomplete. You can still add notes, past Q&A or anything else; nothing is required.

The kind decides the words Whispr uses about the people in the room, what the answers are grounded in, and how a question is classified. A general meeting is never described as an interview and never asks who is the candidate — there is no candidate.

Much of a working meeting is not a question, and Whispr knows that: in a general meeting it only offers an answer when something genuinely calls for a reply, rather than filling the screen with cards for every statement.

Concretely, a general meeting answers what was actually asked — a question mark, or an opening like what, how, when, should we — and records the rest. It deliberately does not treat "tell me about the migration" as a question: in an interview that phrase is the question, and in a team sync it is just how people talk. If you want an answer to something Whispr let pass, type it into the question box, and being addressed by name always counts.

A meeting keeps the kind it was created as. Changing it after material is loaded would leave prospect notes grounding a team sync, and neither is recoverable, so create a new meeting instead.

A meeting's own page

Press a meeting's title, or Open on its card, and you get that meeting on its own page: its details, what you want out of it, and all of its material in one place — with room to read them, rather than a strip inside a card.

What the page asks for is decided by the kind of meeting. An interview asks for a CV, the advert and the research; a sales call asks what you sell and who you are meeting; a general meeting asks for nothing in particular. This is where you add or replace a CV or an advert, change which round it is, and see what is still outstanding.

Each thing the meeting asks for is its own block, and it says what you have: the documents themselves, how much of each is searchable, or nothing yet. A document that was stored but produced no searchable text says so rather than counting as loaded — it grounds no answer. Anything you added that this kind of meeting does not ask for is listed under Also loaded, because it is still searched.

Opening a meeting also selects it, because material belongs to whichever meeting is selected. The one time it cannot is while a meeting is being recorded — moving the scope out from under a running session would change what grounds its answers — and the page says so rather than showing you another meeting's documents under this one's name.

Which round it is

Under the kind is Which round — behavioural, technical, system design, or mixed for an interview; discovery, demo or pricing for a sales call. Leaving it on "not sure yet" is fine and is exactly how Whispr behaved before the setting existed.

It changes how an answer is shaped, not what it is built from. A behavioural and a technical round retrieve the same CV and the same advert — which is why, unlike the kind, you can change this whenever you like. If you find out on the day that round two is systems design, say so then: it is on the meeting's own details, either from Edit on its card or under Meeting details at the top of Prep.

Behavioural is the one worth setting deliberately. Most behavioural rounds are scored against STAR, so that setting tells every engine to make the four parts audible: Situation and Task in a sentence each, the Action in the first person singular — "I did", not "we did", which is the single most common way a STAR answer loses marks — and the Result last, with a number or a date on it. It also forbids reaching for a cleaner example than the ones in your material: if nothing you have loaded fits the question, Whispr says which of your real experiences is closest rather than inventing a better one.

Preparing with a Claude session. get_meeting_brief tells that session which round this is and what it asks for, so the material it writes with you is the right shape — specific past events with outcomes attached, rather than general capability statements. Push it back with add_prep_material and it grounds every answer on the day.