Initiatives
One conversation that runs the whole launch.
Tell an Initiative what you want to happen and why. It asks what it needs to, writes an Outcome Contract and a plan for every part of the company, and runs that plan with the same knowledge of your company every capability reads.
How it works
- 1
Describe the outcome
Start from a sentence or from a company discussion. The Initiative asks only what your workspace cannot already answer.
- 2
Review the plan
Read the Outcome Contract, the specs and the full task tree, then accept it, ask for a revision or reject it.
- 3
It runs the work
Each top-level task goes to the capability that owns it. Builds, designs, sales research and videos report back into the room.
- 4
Release and observe
A closing review says what was delivered, released and measured. Observations keep checking the number afterward.
Outcome Contract
An outcome to answer to
Every Initiative starts by writing down what success means, so the plan has something to answer to and the result has something to be judged against.
- Objective, baseline, user outcome and business outcome
- A hypothesis and a measurable target
- Guardrails, non-goals, rollout and a close condition
Locked plan
A plan you approve once, exactly
Before it starts, the Initiative walks every capability to decide what the outcome needs. Accepting the plan pins the exact version of every task, so nothing runs against a plan you never saw.
- Accept, request a revision or reject the whole plan
- Locked tasks keep their scope; status and evidence stay writable
- A scope change comes back as an exact edit for your approval
Closing review
It reviews its own result
When the Builds and setup work are done, the Initiative answers six questions with evidence before it calls the work finished.
- Outcome, delivered, released, measured, recurring work and follow-ups
- Each answered as met, not met or unclear
- Missing measurements and automations are created during the review
Where it stops for you
Runs on its own
- Researches your workspace and drafts the plan
- Starts one Build for each locked top-level task
- Sets up the observations and automations the outcome needs, paused
- Moves its own tasks forward and writes the closing review
Waits for you
- Accepting the plan before any execution
- Every change to a locked task
- The exact design version a Build implements
- Anything that reaches a customer, spends money or ships a release
What it works with
Reads
- Library
- Working folders
- Business metrics
- Operating rules and skills
- Discussions
Produces
- Outcome Contract
- Locked task tree
- Builds
- Design requests
- Sales records
- Observations
- Automations
- Closing review
Where founders use it
A product launch
Design the screens, build them, prepare sales and the launch video, and measure paid customers, all from one plan.
Fixing a funnel
Start from a number that slipped. The Initiative finds the work across product and growth that should move it back.
Testing a market
Write the hypothesis down first. The plan collects the evidence, and the closing review says what was learned.
How it compares with a project tracker with AI
In the Tablebird launch
It starts as one conversation.
Ava describes the launch once: restaurant groups, one dashboard across every location, 15 paying groups by the end of November. The Initiative asks what it needs to, then drafts the plan for every part of the company, with that outcome written into it.
Follow the whole launchWhat people ask
- What happens when the plan needs to change?
- The Initiative proposes exact edits. Work on the affected tasks waits until you approve or reject them.
- Does the Initiative write code itself?
- No. It hands each top-level task to its own Build session and follows its progress.
- Can I steer it while it works?
- Yes. Send a message mid-turn to steer it, stop the current turn, or ask one turn to think harder.
- Does a finished Initiative prove the outcome happened?
- No, and it never claims to. The closing review reports what was delivered and measured, and observations keep checking the number after release.
- Which models does it run on?
- Your own Claude Code or Codex subscription, or a DeepSeek model on your own API key.
How it connects to the rest of the company
Every capability reads the same Library and reports into the same Initiatives.
Early access · macOS
Run your next launch from one conversation.
We’re onboarding a few founders and small teams each week, one at a time, starting with the launch you’re planning next.
We email you once, when your seat is ready.