Work in Motriz
The co-founder
One durable conversation per product: research, challenged assumptions, decision scorecards, build handoff, and post-ship observations.
Co-founder is the third of a product's four locked destinations — Dashboard, Journey, Co-founder, Build. Selecting it opens the product's one durable co-founder conversation: a persistent partner that clarifies what you are actually deciding, challenges assumptions, runs research, prepares checkpoint scorecards, hands agreed work to Build, and asks how things landed after you ship.
The destination exists only on products that are on the product journey. A plain code product shows Dashboard and Build only — the coding harness is complete on its own, and product guidance starts only when you explicitly start the journey.
One conversation, kept for the life of the product
Each journey product has exactly one co-founder conversation, and it is durable by design: it cannot be renamed, archived, or deleted, and it never appears among the ordinary conversations and build runs in the sidebar's Work list. It is pinned above them in the product navigation. The row renders even before the conversation exists; selecting it creates the thread. Starting "new" founder work — from the sidebar, from Start work, or from a Journey milestone — always reopens the same history. There is never a second thread.
This is why nothing about the product scatters: research, decisions, build handoffs, and post-ship observations land in one place, and each new round of the journey opens with what earlier rounds proved and disproved, not amnesia.
The sidebar row
The row uses a diamond icon, sized like every other destination icon. Orange is reserved for the moments the co-founder is actually working: while a run is active, the diamond turns orange, a small pulsing dot labeled "Run in progress" appears at its corner, and the label shimmers. The counted "N RUNNING" pill belongs to the Build row, not here.
One face, presented by activity
Behind the conversation, Motriz coordinates specialist work — competitor mapping, test design, requirements writing, feasibility checks, and more. You never meet those specialists as separate agents. Everything presents as one face: a step the co-founder runs itself is labeled "Co-founder"; delegated work is labeled "Co-founder · {activity}" — "Co-founder · Mapping the competition", "Co-founder · Designing the tests", "Co-founder · Cutting the MVP scope". The label says what the co-founder is doing, never who is doing it. While it works, the transcript streams plain one-line activity such as "Searching the web", "Running deep research", and "Saving an artifact".
Some specialist capabilities carry a "Premium preview" badge to mark work intended for paid-plan depth. It is not a promise of unlimited managed compute. Current access and managed usage follow your account entitlements; local work and providers you connect yourself remain available when the managed allowance is exhausted.
What the co-founder actually does
Clarifies intent
It interviews with at most one or two focused questions per turn, aimed at the unknown that unblocks the most, starting from the product's actual state — an idea, a repository, a prototype, an active product, or a stalled one.
Challenges assumptions
It is honest, never sycophantic — its own prompt calls unearned agreement "the most expensive gift" it could give you. It never invents traction, pain, pricing, market size, or competitors, and it does not recommend building by default: building is the most expensive test there is.
Runs research
Market and competitor questions are researched, not answered from memory. The conversation has web search and page fetch plus a dedicated deep-research mode that returns cited findings.
Prepares decision scorecards
At a checkpoint, "Ask your co-founder to prepare the call" produces an honest read — evidence for, evidence against, open risks — and a recommendation among Continue, Strengthen the evidence, Reshape, Stop on evidence, and Proceed to build. It is advice, never a decision: you record every verdict, and if you override, "both positions will be recorded".
Hands off to Build
When you record Proceed to build, its scoping work becomes an ordered task list where each prompt stands alone — learning objective, acceptance criteria, constraints. Tasks are only queued when you start the handoff, and the co-founder keeps the agreed context at hand while you build.
Collects the observed result
After a work item ships, a banner reads "Co-founder · Ready to hear how this landed" — "your answer becomes the observation — there is no form". "Talk it through" opens the conversation; the co-founder asks one warm question, and your answer is captured as evidence. Shipping is an observation point, not a success claim: "Too early to call is a legitimate reading."
As you talk, durable records accumulate without a save step: evidence, assumptions, and tests land in the Evidence tab; canonical artifacts like the Idea brief and PRD land in Artifacts; and the co-founder keeps a short brief note on the Journey's current milestone card whenever a turn reframes what to do next.
Opening the conversation
- The Co-founder sidebar row.
- The current Journey milestone's "Work on this with your co-founder →" link (other milestones offer "Work on it together" in their detail), which opens the conversation with the composer prefilled — never auto-sent.
- Start work (⌘N): on a journey-stage or stopped product the view is founder-only, headed "Talk it through with your co-founder"; a building product adds a fourth mode card — Co-founder, "Sanity-check the build against the evidence, or talk it through".
- A shipped work item's "Talk it through" button under Build.
The composer, flowless by design
The composer placeholder reads "Message your co-founder…". Founder conversations hide the single/flow toggle and agent routing pickers — the co-founder is one persona, so there is nothing to route by hand, and flows are never offered here. Model, reasoning effort, and the Fast tier stay first-class, and the full composer surface is available: file attachments, pasted images, product artifacts, and voice dictation. On an empty conversation, a ghost chip tagged "Suggested" offers the product's top next step; clicking it only fills the composer — nothing runs automatically. On an empty conversation without a configured provider, the composer says "Connect a provider in Settings to message your co-founder".
Quick-answer question cards
When a reply asks you something, the co-founder can mirror its questions as a card of tappable answer chips under the message: at most three questions, each with two to five options — concrete answers grounded in this specific conversation, never generic filler. Tapping a chip answers in the option's own words, and a free-text field ("Or answer in your own words…") is offered by default. A card is answerable only while its message is the last in the conversation; after that it stays visible but inert.
The product's default model
The provider, model, and effort chosen on the co-founder conversation are the product's defaults for journey work. The add-product dialog labels this picker "Your co-founder", with the helper text "Becomes this product's default. Changeable anytime from the co-founder conversation." Journey work threads on the same product inherit those choices unless you configure them explicitly. See Providers and models.
Knowledge accretes automatically
On journey products, product knowledge accretes from the co-founder conversation as it happens — there is no setting to turn on. After enough fresh conversation, a background pass quietly folds durable facts into the product's knowledge topics (product, business, customers), running on the co-founder thread's own session; if the conversation adds nothing durable, nothing is written. You see this only as the knowledge status chip on the Dashboard and Journey title bars ("Building product context…", "Context ready"). The conversation, the evidence ledger, and this knowledge are stored in the app's local data folder — see Privacy and safety for where.
How it differs from a Chat conversation
| Chat conversation | Co-founder conversation | |
|---|---|---|
| How many | Any number, listed under Work; can be renamed, archived, deleted | Exactly one per product, pinned; no rename, archive, or delete |
| Who answers | The Assistant role | The founder team, routed per turn and presented as one co-founder |
| Context | The product's artifact inventory | That, plus the founder context pack: starting point, prior rounds, canonical artifacts, open assumptions, the evidence ledger |
| Side effects | None | Evidence, assumptions, tests, brief notes, task seeds, and scorecards persist |
| Session | Its own | The product's default, inherited by journey work |
Both share the same composer and both can show quick-answer cards. For the mechanics of runs — activity, stopping, costs — see Work and runs; for the vocabulary behind milestones, checkpoints, and evidence, see Core concepts.