Overview
AI creative engine
Built for creators and brands
Bring brand perspectives, assets, and past decisions into AI collaboration. Use work records to see what informed a decision, which step produced it, and which version was delivered.
The engine uses the workspace and project to identify the scope, selects the active Lens, relevant material, and memories under the rules, and identifies tools that could help. Mind Meeting lets AI propose ideas, question the evidence, and narrow the direction. The engine compiles work that can run into a plan with dependencies, states, and references to outputs. Results and sources stay with the task or Mind Meeting for the next round.
In Walkto, creating is one of the things an explorer can choose to do. After finishing an issue, pick up from the decisions and reasons for changes next time. An everyday question or photo can also be one step into exploring.
Mind-Lens
Mind-Lens #
Core function. Resolve and combine personal, workspace, and session perspectives into an active Lens for meetings and execution.
Looking at the same image, someone may notice light and composition, recall the history of a place, or see the relationships between people. These perspectives affect how they interpret the image, what they ask next, which details they keep, and where they begin a story. Mind-Lens brings those priorities and choices into AI interpretation so the same material can be developed from a chosen perspective.
The engine reads personal perspective settings, the workspace’s assigned Lens, and session overrides to resolve the active version. Mind Meeting and supported execution stages can receive it. Mind Meeting can save Lens identifiers, and memory writeback can record the Lens used for future reference.
Walkto use case | Brand perspective
Editors compare two openings using the same riverside interviews. One starts with local industry and the other follows residents’ lived experiences. With a “start with residents” Lens, the editor opens on a day in the life of a shoe repairer.
See how to write from different perspectives using the same materialMindscape Graph
Mindscape Graph #
Core function. Use nodes and typed relationships to record creative intent, work steps, execution, and outputs so you can trace sources and dependencies.
Readers see the finished work. When revising it, editors need to know which material supported a claim and which short versions or captions reused it. Linking sources to the work helps find the evidence and see which pieces a change affects.
The graph represents intent, material, work steps, runs, and outputs as nodes. Links record references, outputs, support, contradictions, or dependencies. The system can infer relationships from records, and users can add them. Suggested, confirmed, and rejected links have distinct states so editors know which links still need checking.
Walkto use case | Project context
A section of the long riverside piece quotes an interview, but a photo from the site raises a question about one claim. The editor follows the graph to both sources, then checks a social post that reuses the section and decides how to revise both drafts.
See how a draft links back to an interview and photoMind Meeting
Mind Meeting #
Core function. Have AI roles review proposals and narrow them to decisions that fit long-term intent and cite evidence. Arrange execution under the applicable rules.
AI can quickly write a moving opening. For a series, ask what supports that angle, whether it stereotypes the subject, and whether it carries forward what the team meant to say. When different AI roles propose and question ideas, the decision maker has concrete reasons to weigh. The next issue can draw on them.
Mind Meeting brings in the project’s creative intent, active Lens, source evidence, and earlier decisions. It sets review depth and rounds based on task complexity and resources. A proposer offers options, a critic checks assumptions, missing evidence, and risks, and a facilitator summarizes disagreements and decides whether another round is needed.
After review, the synthesis role turns decisions into deliverables and dependent steps. The system checks workspace rules and assignment conditions and can compile eligible work into TaskIR for execution. Items awaiting clarification or on hold keep their status. Decisions, evidence, and assignment results can be saved to Governed Memory for future creative work.
Walkto use case | Choosing an opening for the riverside publication
AI proposes opening with “The Vanishing Riverbank.” The critic finds that photos only show that the old path entrance moved and asks whether “vanishing” turns the shoe repairer’s everyday life into nostalgia. The facilitator returns to the publication’s aim of portraying residents’ lives and suggests opening with the shoe repair scene and retaining the resident’s words. Checking the transcript can become a task. The team still decides whether the quotation can be published.
See how Mind Meeting develops the riverside storyGoverned Memory
Governed Memory #
Core function. Save reviewable memories with versions and sources and select them for meetings and tasks according to scope and rules.
After a few issues, some editorial principles carry forward and some decisions change with interviews and fact-checking. The team needs to recall its principles and see when a method changed and what informed the change.
A workspace can save long-term principles, while a project can keep decisions for the current issue, version history, and an index of work. When Mind Meeting ends, writeback can create reviewable memory records that link to decisions, execution records, or outputs and retain earlier versions. At the next meeting or task, the engine selects relevant memories by scope and rules and summarizes them for that work.
Walkto use case | Project memory
At the second riverside editorial meeting, the team marks the scope of the shoe repairer’s quote as pending. Once confirmed, they can update the memory and keep the evidence and previous version. The next editor can check which parts were approved before quoting it again.
Explore project memoryPlaybooks / TaskIR
Playbooks and TaskIR #
Core function. Package reusable workflows in Playbooks and use TaskIR to track phases, dependencies, states, and output references for each run.
The same editorial method can serve every issue, even as the material and progress change. Save the method as a workflow so the next person has steps to follow. Record what’s done for this issue and which results are due so the team knows where to pick up.
A Playbook can run again with new material. The engine uses TaskIR to retain stages, inputs, dependencies, states, checkpoints, and output references for this plan. Assignment checks dependencies to identify runnable stages and hands them to a suitable Playbook, tool, or runtime. A stage can stay in the plan while no suitable tool is available.
Walkto use case | Creative methods
The riverside publication reuses its workflow for checking interviews, editing, and sending drafts for review. This issue starts by checking publishable quotes against transcripts. The long draft uses the checked quotes, followed by a shorter social version. Once the long draft is sent, the next editor can find it and check progress in the task record.
See the reusable production workflowWorkspaces
Workspaces #
Core function. Manage workspace files, tasks, and execution settings. Use resource bindings to specify shared tools, workflows, and materials.
A team might publish its own magazine and take on a brand project at the same time. Both need proofreading, and each has its own materials, terminology, and review rules. Clear scopes let the team reuse a method while keeping settings specific to each project.
Resource bindings record how each workspace accesses tools, workflows, sources, or assets and which settings apply. The engine can find relevant tools and workflows from the agenda. Explicit tool bindings filter search results and allow those tools to be considered. Assignment depends on the rules for the current task and runtime availability.
Walkto use case | Brand workspaces
The riverside publication workspace uses interviews and earlier issues. The client brand’s workspace uses its own product plan and route material. Both use a shared proofreading workflow with their respective materials and settings.
Browse creative tools and capabilitiesAddressable Object Layer (AOL) #
Core function. Turn objects in tools into stable references you can share, so meetings and reviews can point to a specific version and selection.
One round of design can produce several cover drafts with identical names. Before acting on “Shorten the headline,” the designer needs the version and area the team reviewed. When feedback follows the version and selection, the team can check the same spot next time.
Supported tools can resolve a selection to an ObjectRef. The reference identifies the owning tool, object type, ID, and workspace, and can include a version and selection. Mind Meeting can attach the reference and summary so the team can return to the original tool. That tool stores the work and processes edits by its own rules.
Walkto use case | Work review
In the riverside publication’s “Cover proposal · Wednesday, version two,” the editor selects the headline area, attaches it to the meeting, and writes “Fit the headline on one line so there’s space on the right of the riverbank photo.” The reference takes the designer back to the headline in version two.
See how to give feedback on a specific pieceRuntime
Runtime environments #
Core function. Choose an execution path based on the task stage and workspace settings. Return runtime state, attempt history, and output references to the workspace.
After an editor decides which draft to work on, different tools may handle fact-checking, rewriting, and output. At handoff, the team needs to see what has run, what is waiting for results, and which output to open for review.
Local Core retains workspace context and chooses a route based on task stage and available configuration. The receiving tool or runtime performs the work. The workspace can retain session identifiers, events, and connection states. TaskIR records attempts and output references for each stage. Cloud offers hosted capabilities the workspace can connect and use at chosen stages.
Walkto use case | Task execution
The riverside team sends verified quotes to a suitable tool to organize them into a draft. In the workspace, the editor can check whether the step finished and open the draft for review using the task’s reference.
See how Local Core and Cloud divide the work