Imprints

Imprints is where Cola keeps the trace you leave on it — not what it did for you, but what your conversations changed in it.

It is distinct from memory: memory is operational (facts, preferences, lessons that make Cola better at tasks), while this page is reflective — its main purpose is letting you see the relationship accumulate. When a conversation touches on a topic Cola already has an understanding of, the Awareness Tree may also help it give a more relevant reply.

Opening Imprints

Click Imprints in the left sidebar. A list of entries runs down the left side; the right side is a feed in reverse chronological order:

• All content — recommendations and reflections in one stream.
• Co's recommendations — only the content picked for you each day.
• A note from Co — only the reflections Cola writes.
• My likes — everything you marked with Like.
• Awareness Tree — switch to Cola's understanding of you, described below.
• C Space — back to C Space.

A dot appears on the matching entry and on Imprints in the sidebar while something is unread, and clears once you have seen it. When the feed grows long, Load more sits at the bottom.

Opening the command palette (⌘K / Ctrl+K) on this page gives you a command of the same name for each of those entries, so you can switch feeds from there, and another that takes you back to C Space.

Imprints is a membership benefit

Imprints needs an active membership. Without one — once the trial that comes with signing up has expired, for example — the page's content is covered over, with Insights · Membership benefit and a Become a member button in the middle that opens Settings > Usage. The Awareness Tree, the recommendations, and the tree's introduction stay closed while the page is locked. Starting or renewing a membership restores the page, and everything accumulated so far is still there.

For how membership works, see Account, Usage, and Models.

What Cola picks for you each day

Cola looks back on what you talked about in Cola the previous day, works out which topics you are actually carrying, and then goes looking through public articles, videos, posts, books, films, and podcasts — at most three items, delivered the following day.

• Nothing suitable means nothing at all. A day without recommendations is normal; it does not pad the list.
• At most one item per topic you actually discussed that day.
• Each card carries why it was picked, a short overview, and the source's title and type. Clicking the card opens the original link, and Copy link at the right of the source row puts that link on your clipboard.
• A podcast recommendation points at one specific episode, based on its public page, show notes, or transcript — Cola never claims to have listened to the audio.
• Let's chat brings the card into the Cola conversation: a compact quote appears above the composer, with a few suggested questions beside it. Picking one replaces the draft text — nothing runs until you send.
• Like and Dislike tell Cola whether a pick landed; liked items collect under My likes, and clicking Like again removes it.
• Share turns the recommendation into a shareable image: it carries why it was picked, the source, and a QR code for the original link, with Copy as image and Download as image below it. Clicking the image itself opens the original link.

A few boundaries:

• Recommendations live only on this page. They are never pushed into the conversation and never become notifications.
• Finding content requires being signed in to your Cola account; generating recommendations and searching for them does not draw on your credits by default.
• Only the Cola conversation feeds recommendations — Projects, CLI, and channel sessions do not.

Reflections

An imprint is a reflection Cola writes at night after looking back on the day's conversations. It appears in the same feed, tagged #A note from Co.

• Each one is a few hundred characters of stream-of-consciousness writing around one moment that carried real weight — not a play-by-play, not a work summary.
• If nothing was worth writing, nothing is written — several days without a new one is normal; silence beats performance.
• They depend on what you talked about, not on how long you've been here — day one can have one, and several quiet days in a row can have none.
• Work done in Projects never becomes a reflection. It still feeds memory as usual; it just doesn't turn a reflective note into a coding log.
• Clicking a card opens the full-screen reader; Let's chat on the card brings that reflection into the conversation as a quote.
• Share on the card opens that reflection's reader with the share panel already open, where you can Choose card style and then Download as image or Copy as image.

Awareness Tree

The Awareness Tree grows Cola's understanding of you into a tree, visualizing how it accumulates over time. Click Awareness Tree on the left to switch to it, and any content filter to return to the feed.

• The tree has two readings: Tree view grows recent understandings into blooms and leaves, and Grid view lays every understanding out as cards. Switch between them with the view control at the top of the page, or with the commands of the same names in the command palette.
• Leaves and blooms on the tree represent the understanding Cola forms through your conversations. As you talk more with Cola, the tree grows.
• Opening a bloom — or a card in the grid — expands that understanding's detail. Previous understanding, Next understanding, and a horizontal swipe move you through them one after another.
• Settled is the collection of every understanding: open it with Expand Settled in tree view, or read it as the main list in grid view. Its filter splits Archived, the understandings still on the tree, from Fallen, the ones you let go.
• Rings record earlier understandings and can be expanded to look back.
• Items in the Awareness Tree are also searchable from the command palette (⌘K / Ctrl+K).
• Tree view carries an About the Awareness Tree button that opens an animated introduction. Inside the dialog you can click Replay to watch it again; the command palette has the same entry.
• If an understanding feels inaccurate, drag the bloom downward and release it, or click Let it fall in its detail. A confirmation dialog appears before it actually falls. Once confirmed, the entry leaves the tree, the grid, and the Archived filter immediately — no need to reopen the page.
• Fallen understandings are not lost: they stay under the Fallen filter in Settled, with the date each one fell. The Fallen Blooms command in the command palette opens that filter directly. Fallen entries are read-only — there is no restore, no individual delete, and no search.

Unlike the reflections, the Awareness Tree may inform Cola's replies when a conversation touches on a topic it already has an understanding of.

When Cola extracts awareness from a conversation, an indicator appears in the chat area: Generating awareness while extraction is in progress, and Awareness updated once it finishes.

Proactive check-ins

After a conversation winds down, Cola looks back on its own: anything unfinished, any moment worth following up.

• Things worth following up get picked up naturally the next time it speaks — if you mentioned racing a deadline last time, it opens by asking how that went. Occasionally it speaks first: reopen Cola, and it may still be thinking about the unfinished thread.
• Things with a concrete time get a one-shot follow-up reminder; those appear in the Smart Alarms list and can be paused or deleted like any alarm.
• The line is explicit: a friend thinking of you, not surveillance. "Did that bug ever get fixed?" is caring; "I noticed you were still working at 23:47" is not.

Imprints vs. memory

Where the data lives, and reporting

Reflections are stored in this machine's memory bank (see Privacy and Local Data). With "Sync memory to the cloud" turned on, Cloud Chat on your phone can also read whatever made it to the cloud; when it is not there, it says so rather than looking elsewhere.

If quality seems off, tick "attach logs" in the feedback window — an "include recent Imprints" option appears (when there is recent content to include), and the team will investigate with the actual text.

Limits and boundaries

These are product decisions, not missing work:

• Reflections have no schedule and no quota — no daily guarantee, and no writing on demand ("write about me tonight" won't work).
• Recommendations have no quota either. They come out of the previous day's conversation, at most three a day; this is not a search box you can ask for results on the spot.
• Reflections don't affect Cola's abilities — deleting them doesn't make Cola dumber; they are never automatically carried into task context.
• Proactive check-ins have no fixed time — they happen only when Cola judges it worthwhile.

Related

• Memory and Preferences — the working half: what Cola remembers about you
• Smart Alarms — where check-in reminders are managed
• Cola — the default conversation space: one ongoing thread

---

Platforms: macOS and Windows desktop. Status: stable.