Skip to content
TakeNext

Research continuity · Private development

Your research shouldn’t restart when the chat does.

TakeNext keeps the problem, evidence, decisions, and history together as work moves between sessions and models. A returned receipt records what ran. It does not decide whether the result counts. The project’s named test does.

Provisional · Cold-read pending
Cross-model workspace not shipped

See what is live today Private access · Sign-in required

Target model
Not telemetry

Target operating model · Not live telemetry Capability attachments are not an execution order.
  • PROJECT IDENTITY · NOT SHIPPED STATUS
  • SHIPPED PRIVATE EDGE
  • SELECTED · NOT SHIPPED TETHER
  • HUMAN / AUTHORITY INTERRUPTION
  • OUTSIDE TAKENEXT · NO EDGE
Source-linked state Shipped private
Source-linked project state and history.
Session context Selected · Not shipped
Conversation-first context compilation is not live.
Visible model routing Selected · Not shipped
Per-turn model selection and switching are not live.
Human interrupt Authority boundary · Shipped surface
A person supplies or changes recorded authority. The current Ballot can bind one exact action.
Shipped executable path Shipped private
Expanded in the bounded-path specimen below.
TakeNext authority ends
Named project test Verdict originates outside TakeNext
TakeNext can record the verdict. It does not supply it. A proof checker, experiment, deployment test, or other project judge decides whether the work counts.

THE SOLID PATH RUNS TODAY WITHOUT THE DASHED STAGES.

A runner receipt establishes that the runner returned a receipt. It does not establish that the result passed the project’s test.

Live bounded path

Live bounded path Capability model · Shipped private · Not a recorded trace
  1. Source-linked state

    The action starts from source-linked project context.

  2. Ballot

    Binds one exact action to a recorded decision.

  3. Approval binding

    Binds that action to recorded authority.

  4. Runner receipt

    A separate runner must claim the work. Dispatch is reported only after its receipt exists.

  5. History

    The outcome remains inspectable.

A runner receipt establishes that the runner returned a receipt. It does not establish that the result passed the project’s test.

Reported verification gates Exactly-once · Hostile-hub · Recovery · Authentication · Responsive

Evidence boundary

Shipped private

Source-linked project state, current Ballot, bounded jobs, approval binding, runner receipts, and history.

Verified in code

Restart-safe manager core, versioned schemas, project and plan state, finite-worker contracts, and typed blockers.

Selected · Not shipped

Conversation-first workspace, context compiler, visible per-turn model routing, model switching, Cutset integration, and full plan management.

Outside authority

Scientific truth and completion. The project’s named proof checker, experiment, deployment test, or other judge decides whether work counts.

Selected operating rule · End-to-end path not shipped

You make one decision. The selected manager is being built to take the next executable step until the named test passes, authority ends, or changed evidence reopens the work.
  • The named test passes
  • Recorded authority ends
  • Changed evidence reopens the work

Refusal

TakeNext does not decide whether its work counts. It preserves state, authority, action, and receipts. The project’s named test supplies the verdict.