Limitless V1 Human Review and Operating Flow
internal prototype · canonical JSON + Dreamborn Forge HTML
internal generated
v1-human-flow-spec · supabase_json

Limitless V1 Human Review and Operating Flow

v1-human-flow-spec artifact · for Limitless OS · status draft

Limitless V1 Decision Brief

Help us validate the operating model before we build the system.

This is not asking you to approve an AI product. It is asking whether this matches how Limitless works today, where the operational drag is, and which human review loops should exist first.

Human Roles
4
review owners
Operating Flows
7
V1 loops
Surfaces
6
minimum UI targets
What We Need From You

Decide whether this feels operationally true.

The useful feedback is not “the AI sounds cool.” The useful feedback is where the model is wrong: wrong owner, wrong review point, wrong first surface, missing source context, or too much friction.

Decisions
  1. Is this current-state map accurate for how Limitless works today?
  2. Where is the most painful operational drag right now?
  3. Who should own each review loop?
  4. Which V1 surface should come first?
  5. What should absolutely require human approval?
  6. What would make this annoying, unsafe, or unrealistic?
How Work Happens Today

Start by correcting this current-state map.

Thoughts and Ideas

Today
Seth has useful ideas in ChatGPT, Claude, conversations, meetings, patient interactions, and daily work.
Likely drag
Ideas are valuable, but they are not reliably captured, routed, prioritized, or turned into reviewed action.
V1 change
Capture thoughts, classify them, and route them into memory, a task, content, a project candidate, or a clarification question.

Front Office Work

Today
Calls, leads, scheduling, cancellations, referrals, admin questions, and patient follow-up depend on human triage.
Likely drag
Betsy/front office has to remember context, draft repeat messages, decide ownership, and manually mark outcomes.
V1 change
Create queue cards with source context, suggested action, draft language, sensitivity, owner, and one-click outcomes.

Content and Growth

Today
Seth has clinical and business insight that could become posts, emails, education, referral nurture, or campaigns.
Likely drag
Ideas often do not become publishable drafts because the capture-to-review path is too manual.
V1 change
Turn captured thoughts into content angles and drafts, then route them to Seth or clinical/privacy review before use.

Clinical and Privacy Boundaries

Today
Some messages touch patient-specific advice, performance data, education, documentation, payment, or PHI.
Likely drag
The team needs a clear line between safe drafting and required human review.
V1 change
Conservatively flag sensitive work, block external use when needed, and provide exact review packets to the right reviewer.

Projects and Improvements

Today
Some ideas are larger than tasks: workflow improvements, SOPs, campaigns, automation rules, or business projects.
Likely drag
Bigger ideas can stay informal unless someone turns them into an owned project with first steps.
V1 change
Promote bigger ideas into project candidate cards with goal, rationale, owner, risks, first tasks, and success criteria.
Proposed V1 Operating Loop

Capture → classify → route → draft → human review → outcome → memory.

1CaptureSeth saves thoughts or the system detects an operational event.
2ClassifyThe system decides whether this is memory, task, content, project, or question.
3RouteThe right human owner gets a queue card with source context and risk flags.
4ReviewThe human approves, edits, rejects, defers, escalates, or promotes.
5LearnReviewed outcomes become reusable memory only when safe.
Discovery Questions

Questions to answer before V1 is considered right.

Today’s Work

  • Where do new leads usually come from?
  • What are the most common front-office tasks every week?
  • What work gets dropped, delayed, or remembered too late?
  • What does Seth personally have to remember that should not depend on memory?
  • What are the most repetitive messages the team writes?

Review Ownership

  • Who should review scheduling and intake drafts?
  • Who should review admin, payment, or insurance language?
  • Who should review clinical-sensitive patient communication?
  • Who should review content before it goes public?
  • Who can approve a new project or workflow change?

Daily Brief

  • Would a daily brief help, or should this start weekly?
  • What are the five things Seth would want surfaced first?
  • Should the brief live in email, dashboard, ChatGPT/Claude, or multiple places?
  • What would make the brief too noisy?

Content and Growth

  • What content format matters first: social posts, newsletters, patient education, referral nurture, or reel scripts?
  • What topics are high value but hard to turn into content today?
  • What tone should never be used?
  • What clinical claims or patient stories are off limits?

Front Office

  • What are the most common missed-call scenarios?
  • What happens today after a cancellation or no-show?
  • What follow-ups should become automatic drafts but not auto-sent?
  • What data would Betsy need on a queue card to trust the recommendation?

Clinical and Privacy

  • What patient messages must always go to a clinician?
  • What education can be drafted safely?
  • What patient data should never leave the source system?
  • What would make the team lose trust in the system?
Seth

Seth

Owns
  • capture thoughts
  • review daily/weekly priorities
  • approve content and brand voice
  • approve project promotion
  • make strategic decisions
  • approve new automation classes
Should not need to do
  • remember every follow-up manually
  • rewrite every draft from scratch
  • search across old chats for operating context
Betsy/front office

Betsy/front office

Owns
  • review front-office tasks
  • send/edit scheduling and intake drafts
  • approve admin/payment language
  • mark outcomes
  • flag workflow friction
Should not need to do
  • manually remember unscheduled eval follow-ups
  • write every repetitive response from scratch
  • guess which items need Seth or clinician review
Clinician/performance reviewer

Clinician/performance reviewer

Owns
  • review clinical-sensitive education
  • review patient-specific guidance
  • review documentation/performance implications when added
  • block unsafe claims
Should not need to do
  • review low-risk admin work
  • hunt for source context
  • approve content with no clinical sensitivity
Privacy/guardrail reviewer

Privacy/guardrail reviewer

Owns
  • review PHI risk
  • patient story use
  • HCS payload safety
  • automation boundary exceptions
Should not need to do
  • review every low-risk content idea
  • manually inspect items already classified as no-PHI/low-risk unless confidence is low
Minimum V1 Surfaces

Pick the first surface that would create visible value.

1

ChatGPT/Claude capture

Seth

  • capture confirmation
  • basic search/retrieval
  • safe prompts for capture categories
2

Daily brief

Seth

  • top priorities
  • decisions needed
  • project candidates
  • content opportunities
  • front-office risks
  • blocked questions
3

Front-office queue

Betsy/front office + Seth as needed

  • task summary
  • source/reason
  • draft
  • owner
  • sensitivity
  • approval actions
  • outcome actions
4

Content approval queue

Seth

  • source thought
  • content angle
  • draft
  • audience
  • risk flags
  • approve/edit/reject/request variants
5

Clinical/privacy review queue

clinician/performance reviewer + privacy/guardrail reviewer

  • risk reason
  • source refs
  • draft/context
  • specific review question
  • approve/edit/block
6

Project candidate queue

Seth

  • goal
  • why it matters
  • source thoughts
  • first tasks
  • risks
  • success criteria
  • approve/defer/reject
Detailed Flow Library

Use these only after the current-state map and decision questions feel right.

01
seth_capture_flow

Seth Thought Capture Flow

Trigger: Seth captures a thought from ChatGPT/Claude using the Limitless MCP.

  1. Seth says capture/save/remember this thought.
  2. Thought Capture Agent writes raw thought to public.brain_memory / limitless.thoughts.
  3. Operating Memory Curator classifies possible outputs.
  4. Chief Operating Brain decides memory-only, queue-worthy, content-worthy, project-worthy, or clarification-needed.
  5. If action is clear, thought_routes and downstream records are created.
  6. If ambiguous, Seth gets a clarification task/question.
Done whenThought is captured and either routed, stored as memory candidate, or converted into a question/task/project/draft.
02
daily_brief_flow

Daily / Weekly Brief Flow

Trigger: Scheduled brief or Seth asks what matters now.

  1. Chief Operating Brain reviews recent thoughts, open tasks, projects, drafts, approvals, and outcomes.
  2. Brief groups items into today, this week, decisions needed, risks, and opportunities.
  3. Brief calls out front-office items, content opportunities, project candidates, and blocked questions.
  4. Seth can approve priorities, defer, ask for more detail, or promote an item to project.
Done whenSeth has reviewed or ignored the brief; any selected actions become tasks/projects/approvals.
03
front_office_task_flow

Betsy / Front Office Task Flow

Trigger: A thought or source-system event creates a front-office task.

  1. Task appears in front-office queue with summary, source, reason, suggested action, and sensitivity.
  2. Relevant agent drafts message/script/recommendation.
  3. Human Approval Router assigns owner: Betsy/front office, Seth, clinician, or privacy review.
  4. Betsy/front office approves, edits, rejects, defers, or escalates.
  5. Outcome is logged: sent, scheduled, no response, converted, deferred, needs SOP, or escalated.
Done whenTask reaches completed, deferred, rejected, escalated, or blocked with reason.
04
content_approval_flow

Seth Content Approval Flow

Trigger: A thought routes to content lane or Seth asks for content from captured ideas.

  1. Content Brain turns thought into angle, audience, format options, and risk flags.
  2. Social Post Drafter or Newsletter Agent drafts content.
  3. Guardrail checks clinical claims, patient stories, PHI, and tone risk.
  4. Seth reviews: approve, edit, reject, request variants, mark as series, or route to clinical review.
  5. Outcome is logged and approved language becomes reusable Limitless Memory.
Done whenDraft is approved, edited, rejected, archived, or routed for clinical/privacy review.
05
clinical_review_flow

Clinician / Performance Review Flow

Trigger: A draft or task contains individualized clinical advice, patient-specific education, testing interpretation, documentation implications, or return-to-sport risk.

  1. Guardrail flags clinical sensitivity and blocks external use.
  2. Clinical reviewer receives source context, draft, risk flags, and specific review question.
  3. Reviewer approves, edits, rejects, or requests more context.
  4. Edited/approved output can return to patient communication/content queue.
  5. Outcome updates Limitless Memory with safe phrasing or boundary rule.
Done whenClinical reviewer resolves the review or marks it blocked.
06
project_promotion_flow

Project Promotion Flow

Trigger: A thought/task implies a multi-step system, campaign, workflow, or initiative.

  1. Chief Operating Brain proposes a project with goal, why it matters, owner, source thoughts, first tasks, risks, and success criteria.
  2. Seth approves, edits, rejects, or defers the project.
  3. Approved project creates child tasks and optional drafts/approval gates.
  4. Project updates appear in daily/weekly brief until completed or paused.
  5. Outcomes become workflow learning or canonical memory.
Done whenProject is approved, rejected, deferred, or converted back into a single task.
07
outcome_learning_flow

Outcome to Limitless Memory Flow

Trigger: A human approves, edits, rejects, sends, publishes, schedules, converts, defers, or escalates an item.

  1. Outcome is recorded against task/project/draft/approval.
  2. System detects whether the outcome teaches a reusable preference, phrase, rule, or anti-pattern.
  3. Low-risk learning becomes proposed Limitless Memory.
  4. Canonical operating principles, sensitive learning, or major workflow rules require approval.
  5. Future agents retrieve Limitless Memory before drafting/routing similar work.
Done whenOutcome is logged and any reusable learning is stored, proposed, or discarded.
Review Actions

Every queue item should have a clear human decision.

approveAccept as written and allow the next safe step.
editHuman changes the output; edits become learning candidates.
rejectDo not use; reason becomes rejection learning if useful.
deferKeep for later with optional reminder/date.
escalate_to_sethNeeds owner/strategy/voice decision.
escalate_to_clinicianNeeds clinical/performance review.
escalate_to_privacyNeeds PHI/patient story/payload safety review.
mark_as_projectPromote from single item into multi-step project candidate.
mark_as_rule_updateTreat as SOP/workflow/automation rule candidate.
Known Open Questions
  • Is Betsy the only V1 front-office reviewer, or should another role be modeled?
  • Should Seth receive daily brief by email, dashboard, ChatGPT prompt, or all initially?
  • Which content review format matters first: social post drafts, reel scripts, newsletters, or content idea briefs?
  • Who is the named clinician/performance reviewer for V1 clinical-sensitive content?
  • Should project candidates be shown immediately or bundled into a weekly review?
Machine Source

The JSON remains the canonical source for agents.

This page is the human decision brief. The full structured payload is available through the JSON link in the header.