Scaffold
Six weeks later, you can still show your working.
Most product analysis is unverifiable by the time anyone questions it, because nobody recorded what it was based on or what has changed since. Here is how that stops.
Join the waitlistOne activity, start to finish.
Each step exists to take one more thing off you: deciding what to do, remembering the context, judging whether the answer is sound, and proving it later.
- 1
Open an activity
The workbook names what is worth doing next and shows the activities its answer will be built on.
- 2
It reads your record, not your memory
Dependencies are declared, so Porter’s reads your actual Competitor Profiles and Market Watch.
- 3
Nine activities get a council
On the nine that make claims about the outside world, three models draft and a judge reconciles. Everything else runs one flagship model.
- 4
You approve the answer
Nothing enters your product record until you commit it. Revise, export, or turn it into a deck.
- 5
Everything downstream knows
Change a persona and the deliverables built on it are flagged as based on outdated context.
AI moment
Never be the person who could not say where it came from.
The fastest way to lose a room is to be asked where a number came from and not know. Every answer carries its own provenance, so that question has an answer before it is asked.
How: the byline names the models that ran it and how long it took. Agreement is marked validated; divergence is named rather than papered over.
It keeps working after the launch.
The product you are deciding on today is one you will still be answering for in two years. The library follows the whole lifecycle, so the record does not stop at launch and neither does the reasoning behind it.
How: 48 activities across four phases, deepest in Developing where the deciding happens.
Deployed carries no live activities yet. One deliverable elsewhere in the model is still to build.
See all 48 activitiesWhat is coming, said plainly.
None of the four below exists yet. They are on the way, and they are the reason some people will want to be on the waitlist rather than wait for launch. Everything else on this page is live today.
Bets
Not yetYour workbook is the slow half: the strategy, and what you know about your market and your customers. A bet is the fast half, one idea you are weighing. Should we build this? Today you argue it from memory. With Bets you describe the idea and Scaffold reads it against your own workbook.
What it will tell you
- How it fits the vision and the strategic pillars.
- Which persona and which job it serves.
- How big it is and what it touches.
- Its riskiest assumption.
- What your workbook could not tell it, and how old the evidence it used is.
Then the loop closes
You shape the bet, decide it, and record why: go, test first, no-go or park. When it ships or is dropped, Scaffold proposes the updates to your workbook, so the strategy you decide against next month is not last year's.
How it will work: evaluate → shape → decide → write back. The seven activities held back from the library come back with it. Six shape one idea at a time (mock-ups, epics, experiments, usability, trial and beta, Kano). The seventh, Features, returns as a product-level register of what the product does today.
MCP server
Not yetThe AI tools you already have open stop being strangers to your product. No pasting context into a chat window before you can ask it anything useful.
How it will work: your workbook exposed over MCP, so an assistant can read the product record directly.
CLI
Not yetFor the product engineers who never leave the terminal, and who would rather script a run than click through one.
How it will work: the same REST API the web app uses, driven from your shell.
The Deployed phase
Not yetThe product after it ships: growth, operations, and keeping the record honest once real users are in it.
How it will work: not settled yet. Re-deciding as the evidence changes is Bets' job, so this phase is about the product in market, not a second pass at strategy.
Be one of the first ten products in the workbook.
Early users shape the activity library, because the first ten are worth talking to. Private alpha, by invitation: no price, no card. Your workbook is stored in the EU; the AI runs through US providers. How your data is handled.