Evidence-led product strategy

Find the focus.
Build the roadmap.

Shinkai gathers scattered product evidence into a shared Focus Map, then builds the map into a roadmap your team can explain.

Map evidence together See collaborators live Keep revision history

From signal to sequence

Every roadmap decision should trace back to evidence.

01 / EVIDENCE

Bring the signals together

Capture needs, requests, market pull, and the advantages only your team can build.

Interview · backlog · research

02 / FOCUS MAP

Make tradeoffs visible

Place each sticky where the evidence puts it. The overlaps reveal focus, gaps, and four familiar product traps.

The focus

03 / ROADMAP

Sequence by evidence readiness

Shinkai auto-sequences Now, Next, and Later from evidence readiness plus impact, size, and dependencies — with an undo. Readiness stays a visible suggestion, ambiguous items wait in Review, and you can override any lane with a recorded reason.

Now

Validated focus

Ship the learning loop

Close one evidence gap

Later

Explore the weak signal

A real working space

Strategy stays connected as the work changes.

The Focus Map is a working surface. Your team can test an idea on it, invite other perspectives, keep a record of what changed, and carry the same evidence into the roadmap.

Focus Map

Name what users want and need, the market pull, and what only your team can build.

Collaborative workshop

Invite a teammate into the same field, with live cursors, shared stickies, and an expiring room link.

History

Save explorations and return to earlier revisions without losing the reasoning behind a decision.

Evidence roadmap

Generate a Now / Next / Later view from the map and export the result as Markdown.

Privacy & control

Sign in to explore. Shinkai saves your work to your account, and it stays private until you choose otherwise. Collaboration happens through explicit invite links to shared sessions; you decide when they start and end. Questions about your data? Contact support@shinkailabs.com.

Data controls

What Shinkai stores. Your explorations and their revision history, uploaded sources with the excerpts parsed from them, backlog items, and your private Ask Shinkai conversations with their citations. Intelligence runs also record non-content metadata — model, timing, and cost — never your prompt or answer text.

Deletion you control. “Delete original” immediately removes a source’s uncommitted parsed excerpts, redacts proposals left without evidence, and queues the uploaded object for durable deletion; excerpts already cited by canonical backlog items remain as provenance. “Delete history” permanently removes every message and citation in a private conversation. After either deletion, only redacted usage and audit ledgers remain.

What reaches OpenAI. When intelligence is enabled, Shinkai sends only bounded, authorized context — never your original files. Shinkai requests no API response storage, but OpenAI may retain API content in abuse-monitoring logs for up to 30 days unless this project is approved for reduced retention. See OpenAI’s data retention controls.

Keys and audit. Provider keys never enter the browser. Every intelligence call runs server-side, within a budget, and is recorded in an audit ledger you can inspect on request.

The center is where the product should live.

Start with one question and one sticky.

The ground it stands on

Where a roadmap comes from.

A product owner works in a loop: gather what users say, decide what it means, commit to a sequence, then defend that sequence to everyone who wanted a different one. The loop runs on a few named artifacts, so the names are worth defining.

A user story is a short claim about value: as a certain kind of user, I want a capability, so that a real outcome improves. The format matters less than the discipline behind it. A story earns a place on a roadmap when evidence supports the want, the need, and the outcome; until then it is an invitation to go find out.

Acceptance criteria define done before work starts. They are the observable conditions a story must meet, written so a teammate can verify each one without asking the writer what they meant. Criteria a stranger can check turn a wish into something testable.

Most tools store these artifacts, and storage was never the hard part. Backlogs grow, stories drift away from their sources, and six months in nobody can explain why one item ranks above another. Shinkai exists for that failure: it pins every story to the evidence that justified it, scores coverage across want, need, market, and build, and turns the coverage into a sequence.

The philosophy underneath is short. Evidence over enthusiasm. A person reviews every judgment, because software can propose and only people can commit.

The Focus Map

Four circles, thirteen regions, one center.

Want, Need, Market, and Build overlap into named regions (user love, trend, feature, commodity) with the focus at the center. The dotted callouts mark what you ship when exactly one circle goes missing.

Thirteen regions, four traps

Every overlap has a name and a consequence. Each dimension is either covered by evidence or still missing it, and the map shows the gap.

Want — what users ask for by name.
Need — the job that costs users hours or money.
Market — willingness to pay, at scale, right now.
Build — what only your team can make.

The four failure modes

Vaporware — Want, Need, and Market without Build: a promise you cannot keep.
The Fad — Want, Market, and Build without Need: attention without a job to do.
Passion Project — Want, Need, and Build without Market: loved, unfunded.
Useful, Not Wanted — Need, Market, and Build without Want: no pull.

Anatomy of one decision

Follow one signal from raw evidence to a roadmap you can defend.

Illustrative example — static sample content, not live product data

Source
A line someone wrote

Every chain begins with a source you can reopen later: an uploaded file, an interview note, an observation from the field.

Sample source line

interviews/2026-05-atlas-onboarding.md · line 42

“We rebuilt the same status report in three different tools last quarter.”
Extracted item
The signal becomes a backlog item

Shinkai lifts a candidate item out of the source and keeps a pointer back to the exact line, so you can reopen the origin of the claim at any time.

Sample backlog item

type: need · source: interviews/2026-05-atlas-onboarding.md, line 42

Unified status reporting — one report that travels with the work instead of being rebuilt per tool.
Human review
A person confirms the theme and the evidence

A teammate accepts the item, names the theme it belongs to, and keeps the evidence attached. Until that review happens, the item stays out of the shared backlog.

Sample review decision

reviewed by M. Sato · 12 May 2026

Accepted into theme “Reporting fatigue” with two supporting quotes and one duplicate merged.
Evidence readiness
Coverage, checked dimension by dimension

Shinkai checks each item against the four dimensions. A dimension is either covered by evidence or still missing it — states in words, never a confidence number.
Sample coverage score
- Want — covered
- Need — covered
- Market — covered
- Build — evidence still missing
Focus Map placement
The item lands where the evidence puts it

On the Focus Map the sticky sits in the overlap of the dimensions it has earned: close to the center, with the remaining gap in plain view.

Sample placement

Placed in the Want · Need · Market overlap, one ring away from the center until the Build question is answered.
Roadmap lane
Shinkai sequences the lane; you can override

Readiness is a suggestion, not a priority. Shinkai auto-sequences each item into Now, Next, or Later using readiness combined with impact, size, and dependencies — with an undo — while the readiness badge stays visible.

Sample roadmap entry

Sequenced to Next — “Unified status reporting” · readiness: one named proof missing · high impact → Next; promote to Now once a build spike confirms feasibility.
Decision history
History stays attached to the decision

Every revision records who changed what and why, so six months later “why is this on the roadmap?” takes one click to answer.

Sample history entry

revision 12 · “Moved to Next after the Atlas onboarding interviews” · saved by M. Sato

Every save is a revision you can reopen without losing the reasoning behind it.

About Shinkai

Evidence over enthusiasm, and a person at every verb.

I built Shinkai (心界, “a world of focus”) on one discipline: every roadmap claim should trace back to something a user said, paid for, or built. Generated ideas now cost nothing, so the scarce work is choosing among them, and a choice you can defend needs its evidence attached.

The second discipline is that people keep the judgment. Shinkai extracts, proposes, and summarizes, but nothing reaches the shared backlog, the Focus Map, or the roadmap until a person reviews it and stands behind it. Placing a sticky by hand is how a team learns its own product.

The thinking comes from my essays on AI, human agency, and product judgment, and the next section walks through the ones that shaped the tool.

From my writing

The essays behind the product.

I write at michaeljabbour.com, and Shinkai is several of those essays turned into a tool. In Of Port and Purpose I describe asking a leadership team which port they were sailing toward and hearing silence. That meeting is why every exploration here begins with a focus statement written in the center of the map.

The Activation Energy of Everything argues that finding and applying patterns has become close to free for anyone with a phone. Once ideas cost nothing, the roadmap question shifts from what a team could build to what the evidence supports, and the Focus Map exists to make that choice visible. The First & Last Principle defines agency as noticing real options, choosing one you stand behind, and acting on it. That definition became the review step in every chain: a person accepts each item, or it goes nowhere.

The Doing Was the Knowing is about what professionals lose when tools take over the handling of their material, so Shinkai keeps the placing, naming, and sequencing in your hands on purpose. The Flip treats hallucination as a property of pattern-matching systems rather than an occasional defect, which is why AI-extracted items arrive as untrusted proposals pinned to their source lines. And Tuning Forks in a Tornado describes a distributed team that had stopped feeling the same moment. The shared map, one field with live cursors and a single center, is my answer to that problem.

Shinkai requires JavaScript to open an exploration. Without JavaScript the Overview tab stays readable; the How it works and About & writing tabs need JavaScript to open.

心 heart · 界 world — you bring the vision, Shinkai carries the craft

You don’t need to be a product manager.
You need to know what you care about.

Drop in everything you have — documents, a repo, a voice note, a hunch. Shinkai is your product manager: it interviews you, drafts the stories, maps the evidence, and sequences a plan. Every judgment stays yours; nothing becomes real until you say so.

Not a generic PM — calibrated to one method Evidence over enthusiasm A person at every verb Show the gap, not the glow The defense is a deliverable

Your next step

The pro surfaces — Backlog, Personas, Workshop, Roadmap, Spec, Decisions — are the same record, one tab away. Shinkai proposes; only you commit.

Who are we building for?

Personas drafted from your sources, refined by you. Every trait is a claim — keep it pinned to something someone actually said, and let the refinement interview sharpen the rest.

Execution source · generated from this exploration

The spec

One page distilled from your sources, map, and roadmap: the hypothesis the plan answers to, the decisions that bind it, the people it serves, and a ranked backlog where every story carries its acceptance criteria and its receipt.

Hypothesis & definition of done

Decision records

Locked decisions bind the plan. Pending decisions sit on the owner's desk. Nothing pending is silently assumed.

The thirteen principles

The rubric everything maps against. Each card shows the pattern that expresses the principle and the anti-pattern that betrays it. “Violates 4 and 9” is a complete design objection.

Research digest

The evidence behind this plan, stated honestly: what the cited sources cover today, and what is still missing.

The people we serve

Projected from this exploration's personas: who the plan serves, and what each of them cares about.

Backlog · ranked, sized, with acceptance criteria

Every canonical item from this exploration, numbered in ranked order. Click a row for the user story and acceptance criteria; edits happen in the story editor and flow back to the roadmap and the map.

Sequencing

The four lanes as they stand on the Roadmap — Now unblocks what follows; Later reaches outward only once the foundations exist. Readiness advises in words, a person commits every lane, and each manual call keeps its recorded rationale.

Workshop

Place each question where it honestly belongs on the map — drag it, or pick one and answer it below.

New sticky

No question selected. Sticky lands at the edge.

Waiting to be placed 0

Committed canonical items. New commits land on the Focus Map automatically — each placed in a suggested region based on its type, with unclassified items waiting in the review strip. Drag any sticky to where it belongs. Use “Place on map” for anything placed later or removed. Manage them in .

The focus

Turn source material into a backlog

Drop in notes, stories, research, or spreadsheets. Shinkai turns them into backlog items and keeps every item tied to its source — undo any of it from the Workshop.

1. Collect sources

0 in the box · parsed independently — one bad source never discards the batch

The collection box. Anything with signal belongs here — documents, repos, typed notes, or your voice.

Drop anything here
docs · decks · spreadsheets · HTML · JSON — Markdown, text, CSV, Excel, Word, PowerPoint Up to 20 files · 5 MB each · 20 MB per batch

Type or dictate a note

or the mic in the top bar

Import a GitHub repo

README and curated docs from a public repo, treated as untrusted evidence like any other source.

2. Set the direction

A short interview before Shinkai writes anything. Pick a box in each row — the answers steer what step 3 produces.

More stories and interpretation cost more per run. Interpretive still cites the source — inferred items are labelled, never invented.

3. Produce the user stories

Shinkai writes the backlog from your sources, shaped by the interview — every item pinned to its source line. Items land staged on the Workshop map, where a banner lets you open the Workshop or undo the whole batch.

Sources stay private to this exploration. “Delete original” immediately removes uncommitted parsed excerpts, redacts proposals left without evidence, and queues the uploaded object for durable deletion; parsed excerpts already cited by canonical backlog items remain as provenance. When intelligence is enabled, creating the backlog sends only bounded parsed excerpts—not original files—to OpenAI. Shinkai requests no API response storage, but OpenAI may retain API content in abuse-monitoring logs for up to 30 days unless this project is approved for reduced retention. OpenAI data controls.

Canonical backlog 0

Every committed item in this exploration, grouped under its epic. Items keep their source links and feed prioritization, the Focus Map, and the roadmap; edits open the story editor and flow everywhere the item is reflected.

Roadmap from the evidence

A Now / Next / Later sequence, not a calendar. Shinkai proposes an order from evidence readiness, impact, size, and dependencies; you commit the lanes, and anything unclear waits in Review.

Select one or more cards, then drag with a pointer or use Space and the arrow keys. Drop anywhere in a lane to move a card there — the insertion line is only for precise ordering. Escape cancels; Space drops. Overriding a lane that disagrees with the readiness suggestion records your toolbar rationale with the move — an empty rationale is recorded as an override without one; your commitments are never overwritten by the sequencer.

North star Write a focus statement in the workshop.

Now

Build now. Shinkai auto-sequences ready items with a priority signal here; readiness stays a visible suggestion and you can override with a reason.

Sequence soon. Auto-placed when one proof is missing or a Now signal is not yet there; readiness stays visible and you can override with a reason.

Later

Hold for later. Auto-placed for exploring or lower-signal work; readiness stays visible and you can override with a reason.

Review

Ambiguous. Items Shinkai will not force-place — thin readiness and no priority signal — wait here with a readiness badge until evidence or a signal arrives.

From your backlog 0

Read-only. Canonical items appear as tiles in the lanes above, each carrying a readiness suggestion you weigh before committing a lane; this section lists only what is not tiled (done or archived items, or overflow past the tile capacity), grouped by status — separate from the evidence lanes above.

Workspace & team

Invite people and set who can do what. Personal and shared work stay separate, and access changes take effect immediately.

New workspace

Choose a workspace

Directory and invitations load only after server authorization.

Invitations

Pending, accepted, expired, revoked, and failed invitations remain visible to team managers.

Shinkai does not send invitation email yet. Generate a one-time link and deliver it yourself through a trusted channel. “New link” is the resend action: it rotates the token and invalidates the previous link.

Invitation lifecycle
Email	Role	Status	Invited by	Created	Expires	Actions

Your explorations

Every saved map keeps its full history. Open one to continue, or start fresh.

Platform administration

Read-only operational view for platform owners. Accounts and workspaces appear as identifiers and counts only — never private exploration or source content.

Back to Shinkai

Service status

Users

Workspaces

Find the focus.Build the roadmap.

Every roadmap decision should trace back to evidence.

Bring the signals together

Make tradeoffs visible

Sequence by evidence readiness

Strategy stays connected as the work changes.

Privacy & control

Data controls

The center is where the product should live.

Where a roadmap comes from.

Four circles, thirteen regions, one center.

Thirteen regions, four traps

The four failure modes

Follow one signal from raw evidence to a roadmap you can defend.

A line someone wrote

The signal becomes a backlog item

A person confirms the theme and the evidence

Coverage, checked dimension by dimension

The item lands where the evidence puts it

Shinkai sequences the lane; you can override

History stays attached to the decision

Evidence over enthusiasm, and a person at every verb.

The essays behind the product.

You don’t need to be a product manager.You need to know what you care about.

Who are we building for?

No personas yet

The spec

Hypothesis & definition of done

Decision records

The thirteen principles

Research digest

The people we serve

Backlog · ranked, sized, with acceptance criteria

Sequencing

Workshop

Backlog items waiting to be placed

Turn source material into a backlog

1. Collect sources

2. Set the direction

3. Produce the user stories

Canonical backlog 0

Roadmap from the evidence

Now

Next

Later

Review

Readiness advisory

From your backlog 0

Workspace & team

New workspace

Choose a workspace

Directory

Invitations

Your explorations

Recent versions

Platform administration

Service status

Users

Workspaces

Session ended

Find the focus.
Build the roadmap.

You don’t need to be a product manager.
You need to know what you care about.