Skip to main content
← What's new
NewImprovedFixed

Memory, shared team work, and a calmer chat

Ampere now follows the rules your team approves in Memory, and everyone in your Organization can see the same sessions and Projects. The chat is calmer too, with results under each answer.

AmpereKnowledge baseScheduled tasksPlatform

Ampere now has Memory, where your team approves the rules, facts, and preferences Ampere follows in every session, and everyone in your Organization can see the same sessions and Projects. This update also brings a calmer chat, results under each answer, and fixes that help long work finish.

What's new

Memory

Memory holds the rules, facts, and preferences your team approves for Ampere to follow in every session.

  • Say it once. When you state a lasting rule, fact, or preference, Ampere follows it for the rest of that session and suggests it as a Memory item. Once someone approves it, every session it applies to follows it. An item can apply to your Organization, a Project, one platform, or one person.
  • Review it in one place. The new Memory page shows what's approved and what's waiting for review, and each suggestion shows where it came from. You can approve, edit, reject, add, or remove items, and undo any of it. A suggestion from someone's own words is approved after 3 days if nobody reviews it, and Owners and Admins can change that.
  • One question when things conflict. If a request conflicts with an approved item, Ampere asks whether to follow the rule, make an exception just this time, or change the rule.
  • Brand memory moves in. Memory's Brand section holds every Brand memory report, for your brand and for other companies you've researched, with your logo, colors, and voice at a glance and your Brand images in their own row.
  • Writing guidance moves too. Your Default writing profile's Additional guidance becomes approved items in Memory.

Your whole Organization shares its work

  • Every session and Project opens to your team. Everyone in your Organization can open every session, including scheduled task runs and monitor follow-ups, and every Project with its files. That includes sessions someone started on their own before teammates joined. If one holds something your whole team shouldn't read, its creator, an Owner, or an Admin can delete it.
  • Who can do what. Owners, Admins, and Members can work in any session or Project. Viewers can read everything and change nothing. Owners and Admins can delete anything, and Members can delete what they created, except a Project that holds teammates' sessions.
  • No more sharing one session at a time. The Sessions list opens on All sessions, with My sessions in its menu, and Manage Access goes away.
  • Scheduled tasks and monitors belong to the team. Owners, Admins, and Members can change any of them. Each one keeps running with the person who set it up, and only that person can change its instructions and inputs, until a teammate takes it over. The previous person gets an email. When someone leaves or is removed, their scheduled tasks and monitors pause. A teammate can take one over: a monitor then resumes, and a scheduled task stays paused until they resume it.

A calmer chat, with results under each answer

  • Less to read. Routine steps are now one quiet line each, and a finished turn folds behind a short summary, so the answer stands out. Your earlier sessions get this too.
  • Results under the answer. Pictures and videos appear as tiles you can open, and other files as rows. In new sessions, a file Ampere delivered and saved shows once, marked with where it was saved.
  • See what Ampere read. A Read saved assets row lists the saved assets behind an answer, with dates when they're known.
  • Clear progress. A running reply keeps a moving status line during pauses, and its timer keeps counting through reloads and tab switches.
  • Real Brand images only. Ampere no longer saves a picture it made or edited as a Brand image. A Brand image now comes from a file you attached or from your brand's own LinkedIn posts.

Choose DeepSeek's effort

DeepSeek V4.1 Flash and DeepSeek V4 Pro used to run at Max on every request. New sessions now offer Low, High, and Max, with High as the default: more thinking than Low, with faster replies and lower credit use than Max.

  • New scheduled tasks start on DeepSeek V4 Pro at High.
  • Earlier choices keep Max. If DeepSeek was already your choice, it stays at Max until you pick another level, and so do scheduled tasks that already use it. A session you started before this update usually keeps running at Max, so start a new session to use Low or High.
  • Some work stays at Max. Monitoring, onboarding, and Free Plan tasks keep running DeepSeek at Max.

Your Project's context on every message

Ampere now uses your Project's summary and context files on every message. Until this fix, most messages typed in the browser went without them.

Long work finishes, and failures are clear

  • Free Plan tasks and Brand memory setup keep going. Answer Ampere's question whenever you get to it, even after an update. If a run stops without its result, Ampere picks it up once by itself in the same session. If that fails too, you see how to retry and get an email.
  • Failures show up fast. If a message can't start, Ampere says so within seconds instead of hanging until you press Stop.
  • Restores protect your work. If a session's saved conversation is missing, Ampere restores it and retries. If a new Workspace can't restore a session's conversation or files, you see a clear message and nothing is sent.
  • Clearer credits. A reply refused before it ran now shows "No credits used".

LinkedIn posts stick to what's true

In new sessions, LinkedIn text, image, and carousel posts and the three Free Plan post tasks no longer invent stories. A person's experience comes only from them or their material, private people stay unnamed unless you say to name them, and every number needs a source. On paid plans, Ampere may ask one question when a story is missing. Ampere also says which kinds of sensitive detail it left out. Paid-plan carousels now come with two other openings for the caption, as text posts already did.

Also new

  • A new app frame. Every page sits in one clean panel. The sidebar's Sessions row is now called Agent, and your list of sessions is inside it.
  • Easier report editors. Brand memory report editors show one section at a time beside a side list, and Find a field works with / or Cmd+F. Lists collapse, reorder by dragging, and delete with undo. Social proof and Company profile reports can now be edited.
  • Project counts. Asset counts on Projects now show only what that Project created, and the Assets tab opens filtered the same way. Nothing was removed: choose Reset controls to see everything.
  • Writing profiles open on your Default profile.
  • Approvals. A scheduled run's approval panel now shows the request's full text.
  • AI Tools. Cards and category chips are real links, so you can open them in a new tab.

Why it matters

Telling Ampere the same thing in every session wastes your time, and so does passing work around your team one share at a time. Memory and shared work remove both. The calmer chat puts the answer first, and DeepSeek's effort levels let you decide when a request is worth Max.

Where to find it

  • In a session: Open Agent in the sidebar. To change DeepSeek's effort, open the menu that shows your model and effort on Home or in a session's composer.
  • In a scheduled task: In Scheduled tasks, create or edit a task, then choose the model and effort under How Ampere runs this.
  • In a report: Open a Brand memory report and select Edit.
  • In Memory: Open Memory in the sidebar. It shows a count of suggestions waiting for review.