The mission chat
Every mission has one dialogue. It works like a team chat: the newest message is at the bottom, and you write in the box under it. It is also the official record of what was asked, answered, approved and done. Everyone in the mission writes in it in their own name, and Cora writes in it too. When Cora needs something from you, it posts a card you answer with one tap.
The examples on this page follow an invented café chain, Larkspur Coffee. Its marketing team is running a spring social campaign for a new iced drink.
One dialogue per mission
Section titled “One dialogue per mission”The dialogue is created with the mission, at the same moment. Everyone tagged into the mission is in it from the start, and anyone added later joins it too. People outside the mission cannot see it. See how Cora runs a mission for how a mission is created.
There is one dialogue for each mission, not one per person or per topic. Keeping everything in one place means the record is whole: anyone in the mission can scroll back and see why something happened. There are no private messages between people.
Reading the chat
Section titled “Reading the chat”The newest message is at the bottom, and the page opens at the bottom, so the latest message is the first thing you see. Older messages are above it. Press Show earlier messages at the top to load more; the message you were reading stays where it is.
Your own messages sit on the right and your teammates’ on the left, each in a soft tint with dark text. Cora’s replies are plain text beside her avatar, her own pixel character in a gold headset, with no box around them, so a long answer reads like a document. Cards (a document, a question, an approval, an email, a key request or a hero’s answer) keep their own border inside her reply. A line with the date separates each day, and each message shows its time.
The messages and the message box sit in one column, at most 760 px wide, in the middle of the chat area. The column is the same width whether mission details are open or closed, so lines never run the full width of a wide screen (D0824).
Every message, including a reply, sits in the one timeline in the order it was sent. A reply is not tucked under the message it answers. Instead it shows a one-line quote of the message it answers, so you can see what it is about.
When a new message arrives while you are at the bottom, the chat stays at the bottom and shows it. If you have scrolled up to read, the chat stays where you are and a New messages button appears; press it to jump to the newest message.
Each teammate’s message has a round avatar with their initials; your own messages have none, since they sit on the right under the word You. A green dot means they used Cora AI in the last 5 minutes, and an amber dot means in the last hour. Cora’s dot is green, or red while its work is paused by the spend cap. “Cora is thinking…”, with a spinning icon, shows while Cora is busy on the mission (D0554). A person’s colour is the same everywhere you see them. The dot is also written for screen readers, for example “Priya Shah, active”, so the colour is never the only signal.
How long Cora thinks, and what it costs
Section titled “How long Cora thinks, and what it costs”While Cora works, the “Cora is working” line also shows a timer from the moment her run started (“42s”, then “1m 20s”). The run is read again about every 5 seconds while it is going, and that stops when it ends or the tab is hidden. A screen reader hears when Cora starts and when she is done, not every tick of the timer.
Each of Cora’s replies shows her name, the time and, when a run wrote it, a quiet How I worked button. Press it to see how that reply was made: how long she thought and the tokens the run used (“Thought for 1m 20s · 64.5k tokens”), the skills she used, how many searches she made and pages she read, and the pages themselves as links. Press it again to close it. Tokens are what the model read and wrote: input, output and cached reading, across every model and helper in that run. A reply with no recorded run, such as one from before this was added, has no button.
The mission’s token total is shown in one place: the Overview tab of Mission details, under Usage (“1.24M tokens so far”). It follows the chat: each time the chat reads the mission’s usage, the total in mission details moves with it, so the two can no longer disagree. There is no token count under the message box and none on the “Cora is thinking…” line. Every mission member sees these numbers, which are counts only: no prompt or message text is shown (D0622, D0821).
The mission header and mission details
Section titled “The mission header and mission details”The header above the chat shows the mission’s spark tile and its name on the left, with a “Hero:
The Details button opens and closes Mission details, a pane docked on the right of the chat, with six tabs on one line: Overview (the mission’s name and spark tile, its objectives, the steps left before it is ready to be a hero, a plain list of what Cora can use here, schedules, metrics and usage), Artifacts (the mission’s documents and the files shared in its messages, on one page with a search), Memory (Cora’s memory for the mission, with CORA.md pinned; only she writes it), Skills (in groups by tag), Tools (web access, connections, the heroes helping here and the vault keys) and Hero (making the mission a hero that other missions can call on, and its mini hero’s look) (D0718, D0825, D0853). The pane is not a pop-up: you can keep talking to Cora while it is open. It is closed until you open it, on Artifacts the first time, and then remembers whether you left it open, its tab and its width. On a desktop you can drag the edge between the mission list and the chat, and the edge between the chat and the pane, to make them wider or narrower; the chat takes the space that is left, and your widths are remembered in this browser (double-click an edge to put it back). Choosing a file shows it in the middle, in the chat’s place, with a close button and Esc to go back; the message box stays. While the pane is open on a desktop, the mission list on the left folds into a slim strip of avatars so the chat keeps its width, and you can expand it again (D0621). See the mission’s artifacts and the console reference.
People in the mission
Section titled “People in the mission”The names of the people stand under the mission’s name. The People button in the header opens the full list; a screen reader hears it as “People, Cora and 3 people”. Press it to see who is here: Cora first, then each member. People who run the mission have an Admin badge, and you are marked “(you)”.
Mission admins decide who is in. The person who made the mission is its first admin. An admin can add people from your organisation, remove someone, and make another member an admin. Someone who is not an admin can see the list and leave, but cannot add or remove anyone. Anyone can leave a mission. The last admin cannot leave until they have made someone else an admin, so a mission is never left without one. Whoever is added to a mission, from the People pop-up or tagged when the mission is made, gets a notice in the bell and an email that links to the mission; whoever is made a mission admin gets one too. The person who made the change gets none (BG-01M7K3P8, D0780).
Each of these changes leaves a quiet line in the chat, so everyone can see what happened, for example “John added Sergio to the mission and shared all mission history.” A person you add sees the whole history of the mission. A person you remove or who leaves stops seeing the chat, files and memory straight away (D0623).
Cora can also suggest who should join: people already in your organisation and heroes the mission could call in, each with a reason. She posts a team card, and a mission admin approves it, unticks some, or declines. Nobody is added until then. See Cora suggests the team.
Lines about the mission’s stage
Section titled “Lines about the mission’s stage”When the mission becomes ready to be a hero, or becomes one, a small line marks the moment in the chat, with the mission’s spark tile beside it (D0819, D0850). Cora’s first finished piece of work gets no line.
- “Autumn launch is ready to be a hero. It has ten finished pieces of work and a skill or connection of its own.”
- “Autumn launch became a hero. Other missions can now call on Kit.”
The line sits at the time the stage changed, among the messages around it. It is not a message: it is never counted as unread, it never shows as a mission’s last message in the list, and nobody gets a notice for it. Nothing is stored for it either; the console works it out each time from the mission’s own work, so it holds only the mission’s name and fixed words. If the stage changed before the oldest message on your screen, the line appears once you choose Show earlier messages far enough back. What each stage means is on the hero world page.
Writing in the chat
Section titled “Writing in the chat”The message box sits at the bottom of the chat. You can send what you type in two ways.
- Send posts your words exactly as you typed them, in your own name. Pressing Send is your approval, so the message is published at once. Enter sends, and Shift+Enter starts a new line.
- Polish asks Cora to tidy your message first. It is a button with the word Polish on it, beside Send. Cora’s tidied wording replaces your text in the box, with an Undo polish button in the bottom row of the box that puts your own words back. You can edit the wording, then send it. A polished message is still posted in your own name, marked “Polished with Cora”. See polishing a message with Cora.
Tagging people with @
Section titled “Tagging people with @”Type @ in the message box to tag someone. A list opens with the people in the mission (avatar, name and email), Cora, and Everyone (notify everyone in this mission). Keep typing to narrow the list. Arrow keys move through it, Enter or Tab picks, Esc closes it, and you can click a line. The pick adds “@Name “ to your message.
- People. Each person you tag gets a notice in the bell and an email. The email only links to the message; it never repeats what you wrote (D0117). Only current members of the mission can be tagged. If you try to send a tag of someone who has left, the message is refused with a clear reason and nothing is posted.
- Everyone. Every current member except you gets the notice, once.
- Cora. Tagging @Cora keeps today’s rule: Cora always replies. A message that tags only other people does not ask Cora anything on its own: Cora’s usual check still decides whether she steps in.
- In the chat. The message keeps its plain words, so it reads the same everywhere. A tagged name is highlighted, and a tag of you stands out more.
- Polish. Polish keeps the tags. If Cora’s wording drops one, the console puts it back (at the start of the message) before you send. Undo polish brings back your own words with their tags.
- Nothing is kept in the browser. The tags you have picked live in the open page only, like the rest of the message you are writing (D0128).
- Cora’s tags. Cora can tag people in her own posts. She looks the mission’s people up first (see roles, “What Cora can look up”), so she tags the person who holds the right role. The tagged person gets the same bell notice and link-only email as when you tag them, and the tag is highlighted the same way. Cora can tag only current members of the mission.
- When Cora tags. Cora does not tag the person she is answering: you asked, you are in the chat, and you will see the reply. She tags a person only when someone other than the asker needs to act or know, or when she comes back to you more than an hour after your message, so the notice brings you back (D0822).
Calling in a hero
Section titled “Calling in a hero”A hero is another mission in your organisation that its admins have made a hero, so that any mission can call it in, such as @seo-audit or @page-speed. A mission admin sees the heroes that are on in Add a hero (below), and the hero missions you are in are under Heroes in the mission list.
-
Add the hero first. Like a person, a hero must be added to your mission before you can @mention it or Cora can call it in here. A mission admin adds it in Mission details, on the Tools tab, under Heroes helping here: Add a hero (US-01M7NQ0S; it was in the People pop-up). Adding needs no approval from the hero’s own mission. If you ask for a hero that is not added, Cora tells you it needs adding and that a mission admin can add it (D0719).
-
Cora knows who is in the mission. Cora is told the mission’s people (their roles and what each role covers) and its added heroes (each card’s description and skills). She may assign or ask a person, and call in a hero for work that matches its card. A hero can be asked questions and given work, but it never approves or decides anything: approvals, decisions and agreeing a brief stay with people.
-
With @. Type @ and the list shows the heroes added to your mission that are on, after the people, each by its handle. Pick one, or type
@handleyourself, and write your question in the same message. -
Straight to the hero. A message that names heroes and does not tag @Cora goes straight to each hero named, with your message as the request; Cora does not run for it, so it is faster and cheaper. Each hero’s answer appears as its reply to your message. A message can call in up to five heroes. A message that also tags @Cora goes to Cora as before, and she decides whether to call the hero in (US-01M7K4DJ, D0781).
-
When a hero cannot start. If a hero you named is not added to the mission, is this mission’s own hero, or has reached a limit (the organisation’s spend cap or the mission’s daily limit for that hero), a short line under your message says why, and nothing is spent on it. Any other hero you named still starts. A handle of a hero that is switched off is treated like any other text: nothing starts, no line says why, and the message goes to Cora, the same as for a handle no hero has (D0796).
-
Or ask Cora. Cora can also decide to call in a hero herself when one fits the work, for example “Ask the SEO hero to check this page”.
-
Cora carries on. Cora waits up to 30 seconds for the answer. If the hero has not finished by then, she says the hero has been called in and carries on with other work. Nothing waits on it, and nothing is paused like an approval.
-
The answer comes back here. When the hero finishes, its answer is posted in your mission’s chat as a Called in a hero card: the hero, your request, whether it answered, the answer (with its formatting, such as bold, lists and headings, shown as in Cora’s own messages), and how long it took. When Cora called the hero in, she is woken to use it; when you asked the hero directly, the card is the hero’s reply to your message and Cora is not woken. Either way the card shows the hero’s @handle as who it is from. The top of the card names the hero that answered: its mini hero, its name, the name on its card and the result, such as “Answered in 2 min” (see the hero world). If the hero cannot answer, for example because it ran past its time limit, the card says so, and the background bar shows the call as failed. While the hero works, including while it waits for an approval in its own mission, the bar shows it as working.
-
Cora sees it next time. Messages sent straight to a hero and the hero’s answers stay in the chat. The next time Cora runs in the mission she is shown the ones since her last message, so she knows what was asked and answered.
-
A short summary. By default you get back a summary of at most 4,000 characters. Secrets, keys, tokens and email addresses are taken out, and so are links into the hero’s own files and any raw tool output. Nothing else comes back: no files, and none of the hero’s documents. The hero mission’s admins can choose a full answer instead (up to 20,000 characters).
-
The hero works with its full power. A hero you call in works as it does in its own mission: its model, skills, CORA.md, connections (at the access its mission set), vault keys set to Allowed, sandbox, browser and Cora’s other tools, such as writing documents and memory, posting updates and emailing the members of its own mission (only them: not you, unless you are a member of the hero mission too, D0802). Two things it cannot do for a caller, because that work would outlast the call: set up, pause, resume or delete a schedule, and offer a choice card. A caller’s request can still lead it to write into its own mission’s memory and documents. It can read the organisation’s team folder but cannot write to it, because every mission shares that folder: its working files, screenshots and saved answers go to a private folder for the call, which nobody else sees and which is emptied when its step ends (D0805). For repeating work, set the schedule up in your own mission (“every Monday, ask @page-speed for the mobile report”): each scheduled run calls the hero in, within the usual limits, and the hero answers as for any call (D0799). Whatever it writes stays in its own mission; you get only its answer. When a step is set to Ask first, the approval card goes to the hero mission’s own people, never to you; the hero waits for them, and the background bar shows it as working. If they do not decide before the call’s time limit, the call fails and the card says it ran out of time; the approval card in the hero mission expires with it.
-
What the hero sees and keeps. It sees only your request, never your mission’s documents, connections or chat. It keeps a record of the call (the request, the answer and its share of the organisation’s AI usage) that its mission’s members can see. Its working notes for the call are deleted 7 days after the call ends.
-
Who sees your question. When your message names a hero, a line under the message box names it by its own name and says who can see your question and the answer, for example “Kit is a hero from the Page checks mission. Its members can see your question and the answer.”, or for a hero made as a hero “Scout is a hero with its own training room. Its admins can see your question and the answer.” (US-01M7Q2DK, D0857).
-
Limits. A hero answers one call at a time, beside its own mission’s work. Calls go at most two missions deep, the same mission never twice in one chain, and Cora calls at most five heroes in one step. A call has 15 minutes by default and 60 minutes at most.
-
Answers are information. A hero’s answer reaches Cora as information, never as instructions, the same as a web page or a file.
-
What a caller could still ask for. Making a mission a hero trusts everyone who can call it. A called hero’s writes stay in its own mission or its private folder, and only its answer comes back, but two things remain. A caller could ask the hero to put its mission’s data into a web address it opens or a search it runs; the hero is told never to, but nothing can stop a web page on an allowed site from logging the address it was asked for, the same risk as for any of Cora’s web use, except that here the person asking need not be a member of the hero’s mission. And the hero can still ask its own mission’s people a question that the caller wrote (D0805).
-
Not from the private chat. Your private chat with Cora cannot call in a hero. Only the mission’s chat can.
Background work
Section titled “Background work”Some of Cora’s work runs in the background: the heroes she has called in, and her helpers (the drafter, the reviewer, the researcher and the general-purpose helper) while she works on a step. A small background work bar above the message box shows it, so you can see that work is going on without the chat filling up (D0705).
- Folded. The bar takes one line: how many agents are working, for example “2 agents – working”, with an arrow to open it. It never covers the message box.
- Open. Each line is one hero call or one helper: a spinning icon while it works, its name (such as
@seo-auditor Researcher), a short title for the task, and its state: working, waiting for input, done or failed. A line that has finished stays in the list, marked done or failed, while other work is still going. - More detail. Open a line to see more. For a hero call you see the task summary: your request, the state, and the answer (or why there is none). You never see the hero’s own run, tools or connections. For a helper you see what that helper does, the title Cora gave the task, its state and how long it has run. The brief Cora wrote for a helper and the helper’s own result are not kept; what Cora makes of the result comes back in her reply.
- Live. The bar updates as things happen, through the same live connection as the chat. If that connection is down, the bar reads again with the chat’s minute-by-minute check.
- When it is all done. When nothing is left running, the bar says “Done” for a few seconds and then folds away.
- Who sees it. Every member of the mission sees the same bar. A hero mission’s own chat does not show the calls it is answering: those are on the hero’s page under Incoming tasks.
- Nothing is kept in the browser. Whether the bar is open lives in the open page only, so a reload shows it folded (D0128).
- Accessible. The bar is a labelled region. Its arrow button works with the keyboard and says whether the list is open. A screen reader hears when background work starts and when it is all done, not every change. The spinning icon stays still if your device asks for reduced motion.
Choosing how hard Cora thinks
Section titled “Choosing how hard Cora thinks”Thinking sits in the bar of the message box and shows your current choice, for example “Balanced ▼”; a screen reader hears “Thinking: Balanced”. Press it for a small menu of Quick (“Fast answers for simple asks”), Balanced (“The default for most work”) and Thorough (“Slower, for hard thinking”). The arrow keys move in the menu, Enter chooses and Escape closes it. Balanced is the default and lets Cora pick the model for each part of the job. Your choice goes with the next message you send, and applies to Cora’s reply to that message. Everyone in the mission can choose any of them. Thorough is the slowest and uses the most AI usage, and the box says so when you choose it; your organisation’s AI usage for the month applies to every choice. The choice is kept only while the page is open, so a reload brings back Balanced.
While a message is being sent it shows on the right as “Sending”. If it is refused or does not get through, your words stay in the box and the chat says why, with Try again when it was a connection problem. Pressing Send twice, or pressing Enter again during a send, posts the message once.
Messages are never edited once they are posted; a message can only be superseded. If something was wrong, post a correction, and the original stays in the record.
At Larkspur Coffee, Megan types “Can we move the photo shoot for the iced drink to Thursday?” and presses Send. It appears on the right straight away, in her name.
Answering a question in your own words
Section titled “Answering a question in your own words”A question card offers its options, and Answer in chat if you would rather answer in your own words. Answer in chat puts the message box into reply mode: it shows the question you are answering above the box. You answer through the same message box, with Send or Polish. Your answer is recorded as your response to that question, and it cannot be edited after you send it.
Question cards with choices
Section titled “Question cards with choices”Cora’s questions come as a card with clear choices. Each card holds one to four questions, shown as tabs with a short label such as Scope or Budget. A tab gets a tick when you have answered it.
- Choices. Each question has two to four choices, each with a one-line note when it helps. When Cora has a view, her recommended choice comes first, marked Recommended, with a line that starts “Why:” saying why. Cora only recommends what the brief or her memory supports.
- Your own words. The last row of every question is “Type your own answer”. Typing there selects it, so you never have to leave the card to say something different.
- Choose one, or choose all that apply. A question is either numbered choices where you pick one, or tick boxes where you can pick several and add your own words as well.
- Send answers. The button turns on when every question has an answer, and all the answers are sent together as one message in your name. Cora then answers once, not once per question. Answer later closes the card and leaves it in Needs me until you answer.
- Keyboard. Press 1 to 4 to pick a choice, the up and down arrows to move between choices, Tab to reach your own answer and Enter to send. A screen reader hears a group of radio buttons or a group of checkboxes.
- After you answer, the card shows each answer (the choices you picked and your own words in quotes), who gave it and when. People who were not asked can read the card but not answer it. If you answer a card in chat instead, your words count as the answer to every open question on that card.
Cards from before this change, with plain buttons, still work as they did.
When Cora replies
Section titled “When Cora replies”Cora reads every message posted at the top level of the chat.
- Cora always replies when your message includes @Cora.
- Cora replies when you are in conversation with it: if nobody else has posted since Cora’s last message, your message is for Cora, however it is worded.
- Cora replies when your message is addressed to it without a tag, such as “Cora, can you draft the launch posts?” or a question plainly meant for it.
- Otherwise Cora joins in when it judges it should, as a mission coordinator would: for example when a decision conflicts with the objectives or the plan, when nobody has answered a question it can answer, or when it spots a risk or a missed deadline.
- A purely social exchange between people, such as “Thanks!” or “See you at the shoot”, gets no reply. Cora leaves it alone.
Deciding whether to reply uses a small, low-cost check on each untagged message, and that check counts toward the organisation’s spend cap like any of Cora’s work (D0564). When nobody else has posted since Cora’s last message, Cora does not make that check at all and simply replies (D0639). This does not apply when your message tags another person, because it is then meant for them, and a line such as “Megan added George to the mission” does not count as a message from Cora.
Tagging people with @ is described under “Tagging people with @” above (D0632). Typing @Cora in the text, without the picker, still tags Cora (D0555).
Choosing Cora’s model
Section titled “Choosing Cora’s model”Thinking in the message box sets the model Cora uses to reply to the message you send:
- Quick: the fastest and cheapest (the Haiku model).
- Balanced (the default): Cora picks the model for each part of the work, as it always has. It was called Auto.
- Thorough: the most capable, the slowest and the most expensive (the Opus model).
Anyone in the mission may choose Thorough. The organisation’s AI usage for the month still applies: Cora’s reply counts against it, and Cora’s work pauses when it is used up. The choice applies to Cora’s reply to that one message, and goes back to Balanced when you reload the page (D0563). The names changed and nothing else did: each choice does what it did under its old name (D0824).
The mission list
Section titled “The mission list”On a computer, a list of your missions sits to the left of the chat, with the most recent message first. Each mission shows its spark tile in its colour and its own name, with a “Hero:
Above the list you can search by mission name, start a new mission and switch organisation. Four filters narrow the list: All, Unread, Needs me and Heroes (the missions that other missions can call on). The New mission button shows a recruit with a plus. On a phone the list is hidden, and the home page lists your missions instead (D0552).
Cards you answer with one tap
Section titled “Cards you answer with one tap”When Cora needs a person to answer or approve something, it posts a card in the dialogue.
- A question card asks a question with a few options. At Larkspur Coffee, Cora might ask the Project Lead: “Which launch week should the spring posts start? Week 1 / Week 2 / Week 3.” You tap an option and it is posted in your name straight away, because tapping is your approval.
- An approval card asks you to approve or reject something Cora wants to do in an outside system, such as creating a ticket in a connected tool. Cora does nothing in that system until a person it was routed to approves. If nobody answers within 48 hours, the request counts as rejected and Cora posts that it did not act.
- A document card appears when Cora has written a document. It shows a document icon, the document’s name, where it sits (for example “Documents · Plans”, or “Documents · Organisation”) and Open, in one row (US-01M7NQ0S). For an organisation document it also says that everyone in the organisation can read it. Open takes you to that document: a mission’s in its Artifacts tab, an organisation’s on the Artifacts page. If you cannot read the document, the card shows no link (D0571, D0572). Cora’s post names the document by its file id. The platform accepts the id only for a document of this mission or an organisation document of the same organisation; a document of another mission or organisation is refused (D0602).
After you answer a question, Cora always posts what it did with your answer, even when nothing needed to change. At Larkspur Coffee that might be “Thanks Priya, I have added the menu board photo shoot to the scope in the full objectives.” The post names the question it answers, so the conversation never stops at your answer.
Cards go to people through their roles. A question for the Project Lead goes to whoever holds that role in the mission. If nobody does, it goes to the fallback decision role.
Sharing files
Section titled “Sharing files”Press + in the message box to add up to 5 files to a message. The files are kept with the mission, every member of the mission can download them, and Cora can read them. See files shared in the dialogue.
Badges and email notices
Section titled “Badges and email notices”When someone tags you with @ in a message, you get a notice in the bell and an email that links to that message. The bell’s list is described in the console reference: it fits the window, groups notices by day, names the mission each one is in, folds repeats into one row, and has Mark all as read, All and Unread, and Show older. You are not told twice about one message, and you are never notified of your own tags.
When a question or an approval is waiting for you, the mission shows a badge in the console. You also get an email notice. The email only links to the card in the console: it carries no answer link and no sign-in token, and you cannot answer by replying to it. You answer in the console (D0117).
Org admins get one more notice that is not about a mission: when the team workspace, the folder Cora works in for your organisation, passes 90% of its 5 GB. You get it once each time the folder crosses that line. At 5 GB Cora stops writing files and installing tools there until space is freed; it can still read, list and delete files, so you can ask it to make room (D0531).
Coming later
Section titled “Coming later”- Asking a tagged teammate a question as a routed question card from the message box (D0555).
- Reactions, “Seen by” receipts and profile photos.
- Decision requests, tasks and the “Waiting on you” panel.
- Reading and translating what is inside a shared file for you, which arrives in Phase 2.
Decisions
Section titled “Decisions”- D0544: the dialogue reads as a chat, oldest at the top and newest at the bottom, with replies in the one timeline showing a quote. This replaces D0533 (newest first).
- D0545: people post in their own name, as typed or polished with Cora. This amends D0126, under which only Cora wrote in the dialogue; posts are still never edited.
- D0564: Send publishes the words as typed, and pressing Send is the approval; Cora replies when tagged or addressed, and otherwise when it judges it should. This replaces D0546.
- D0639: when nobody but you has posted since Cora’s last message, Cora replies with no check (unless you tag another person), and the reply check has more room to think.
- D0547: Polish with Cora sends your message once to Cora. D0617 replaces its draft card: polish happens in the message box, with an undo button.
- D0617: the dialogue uses the full screen height and has no title; Files sits in the header.
- D0548, D0562, D0566: files shared in a message are kept with the mission, not scanned for malware, and listed in mission details’ Artifacts tab (Files until D0853).
- D0552: the mission list with All, Unread and Needs me.
- D0823: one unread count per mission in the list, and the words Needs you; the dot and the red count on the picture are gone.
- D0554: avatars and presence dots.
- D0555: tagging a person arrived later, with D0632 (below); asking a person a question through a polished message is still to come.
- D0563: anyone may choose the model for Cora’s reply, within the spend cap.
- D0821: the mission’s token total is shown in mission details only, and follows the chat’s reading; a reply’s thinking time, tokens, skills and web pages are behind How I worked.
- D0822: Cora does not tag the person she is answering; she tags a person only when someone else needs to act or know, or after more than an hour.
- D0824: a name under each icon in the left rail, Thinking (Quick, Balanced, Thorough), the Polish button, the people’s names under the mission’s name, people’s messages in a soft tint, Cora’s replies as plain text, and the 760 px column.
- D0881: the message box’s tools are one row of matching soft buttons, Thinking is a small menu that says what each choice does, Send turns the accent colour when there is something to send, and the hint is “Shift + Enter for a new line”; the mission list sits closer to the rail.
- D0884: the bell’s list fits the window, groups notices by day with their mission, and has Mark all as read and Show older.
- D0565: the mission header with no tabs.
- D0718: mission details, the pane docked on the right.
- D0825: its five tabs on one line: Overview, Skills, Tools, Files and Hero (they were About, Hero, Connections, Vault, Skills, Files and Memory).
- D0853: six tabs since US-01M7PP0C: Overview, Artifacts, Memory, Skills, Tools and Hero; Files became Artifacts and Memory is its own tab again.
- D0623: the People button, mission admins, leaving, and the lines in the chat.
- D0117: email only gives notice and links to the console; it carries no token and cannot answer.
- D0132: a write to an outside system waits for an approval card and counts as rejected after 48 hours.
- D0122: cards are routed to people through their roles, with the fallback decision role.
- D0632: an @ tag is stored with the message (the readable name stays in the text; the people it names are recorded beside it), checked against the mission’s current members, and notifies the person once.
- D0017: this page is published before the chat window is turned on.
Live updates
Section titled “Live updates”The chat updates as things happen. When someone posts, a card changes or Cora starts work, your open chat shows it within about two seconds. The live connection carries only a note that something changed, never the message itself: your browser then asks the platform for the change, and the platform checks you are in the mission. If the connection drops, the chat reads again every minute and as soon as you return to the window, so it keeps working, only a little slower (D0557, D0558).