//the evolution of an idea, made physical·a kanban for thinking, not just tracking

Ideas don't arrive finished.

Ideafy gives them somewhere to evolve — a column for every stage of the thinking, not just the doing. It reads your project, argues with weak ideas, plans the strong ones in plain language, and keeps the receipts when an idea doesn't make it.

download for macOS read a real refusal →

just want to see it? → for founders · for developers

kan-114.md

narrative.md

opinion.md

worktree: kan-114

# KAN-114 · column: Ideation

id: "kan-114"

title: "Ship offline-first sync engine by Friday"

title: "Manual board export/import (.ideafy file)"

scope: one-file, drag-drop, no-accounts

estimate: ~180 LOC · 4 tests · 1 afternoon

verdict: "unknown"

verdict: "refuse original · counter-propose"

reasons:

- narrative.md line 42: multi-device sync is non-goal for v1

- 11 users, 0 sync complaints logged

- CRDT + E2EE cross-platform ≈ 6–10 weeks, not 4 days

awaiting: human sign-off

# no code has been written. no branch has been touched.

§ 01card anatomyfour lenses on the same idea

A card is not a ticket. It's a stable surface where four kinds of thinking sit side by side.

Detail describes. Opinion judges. Solution plans. Tests verify. Collapse them into a single “description” field and the thought collapses with it. Keep them separate and the card becomes legible — to your future self, to a teammate joining tomorrow, to an AI agent right now.

kan-114Manual board export to a single .ideafy filein progress

Detail

What is this?

Description, constraints, context. The idea as given, before anyone has had an opinion about it.

Opinion

Is it worth doing?

Evaluation against goals, non-goals, risks. The verdict — with reasons you can argue back at.

Solution

How will we do it?

Plan, sequence, files touched, trade-offs. The blueprint, written before the first commit.

Tests

How do we know it's done?

Acceptance criteria, written before the finish line so “done” is a fact, not a feeling.

+ chat history per tab·+ git worktree on its own branch·+ merge commit when shipped

A real Ideafy card opened to its AI Opinion tab, showing the Detail / Opinion / Solution / Tests tab strip and a verdict written by the model. — Fig. 1A real card, mid-thought. The four lenses sit in the tab strip; each is its own conversation with its own chat history. The verdict on the right is the Opinion tab writing itself in plain language.

§ 02chain of thoughtkan-114.transcript

This is what happens when you file a bad idea.

Not “> Planning solution...” log spam. A real conversation — the model reads your narrative, refuses the scope you wrote down, counter-proposes something smaller, and waits for you to approve or override before a single file is touched.

kan-114 · ideation → planning · session #a41fawaiting operator

operator14:02:11

New card in Ideation. Title: Ship a full offline-first sync engine with CRDT merge + end-to-end encryption by Friday.

It should work across web, iOS, Android. Also add a conflict UI so users can pick which version wins.

ideafy14:02:14

Reading narrative.md before I answer.

Audience: solo builders and two-person teams
Stack: Next.js, SQLite, single-node deploy
Non-goals (explicit): multi-device sync in v1, team collaboration UI
Current users: 11 active, 0 reported sync complaints

ideafy14:02:16

scopeyagniI am going to push back.

The narrative says multi-device sync is a non-goal for v1. You have 11 users and zero sync complaints in the log. A CRDT engine plus E2EE across three platforms is roughly 6–10 weeks of disciplined work, not 4 days. Shipping it by Friday means shipping something broken that you will spend the next month apologising for.

ideafy14:02:18

Here is what I think you actually want. Tell me if I am wrong.

The pain is: “I opened my laptop at a cafe and my board was empty.”
That is an export/import problem, not a sync problem.
One user asked for it in Discord on Tuesday. One.

ideafy14:02:19

counter-proposalSmaller card. Ship this instead:

KAN-114 — “Manual board export to a single .ideafy file, drag-drop import on any machine.”
Scope: one file format, one button, no accounts
Estimate: 1 afternoon, ~180 lines, 4 tests
Unblocks the real pain. Buys you 6 weeks to see if sync is actually a top-3 request.

ideafy14:02:20

I have drafted the plan and the tests. I have not written any code and I will not until you say so.

Options: [approve smaller card] · [keep the original, override me] · [kill the card entirely]

operator14:03:47

Approved. Go with KAN-114. Start in a worktree, run the tests, stop at the diff.

§ 03preservationthe card outlives the verdict

A refused card is not deleted. It's filed.

Most tools treat a “no” as the end of the line. Ideafy treats it as the start of an archive. Withdrawn cards keep their reasoning, their conversation, their counter-proposals — so when the same idea walks in three months later wearing different clothes, you already have the receipts.

kan-098

Offline-first sync engine, CRDT, E2EE

filed apr 14

withdrawn

→

kan-114

Manual export/import (.ideafy)

shipped apr 14 · 1 afternoon

shipped

…

kan-203

Multi-device sync, narrowed scope

+99 days · scoped from kan-098

backlog

Three months later, the same idea walked in dressed differently. The receipts were already there.

The Ideafy board, light theme, with the Withdrawn column visible on the far right next to Completed. — Fig. 2The board, real. Withdrawn is the column on the far right — not a trash can, an archive. Every refused idea keeps its Detail, its Opinion, its plan, and its conversation, in case it walks in again.

§ 04five movementsspec.sheet

From a thought to a merged diff, in five deliberate steps.

Nothing happens behind your back. Every stage produces a file you can read, edit, or throw away.

Read the narrative

On first run, Ideafy reads the repo and writes a short product narrative — problem, audience, stack, non-goals. Every new card is measured against that document.

narrative.md — yours, editable.

Argue before planning

Ideafication pass runs YAGNI, feasibility and scope checks. Weak ideas are refused in writing. Strong ideas earn a plan. Sometimes the verdict is: don't build this yet.

Refusals are logged, not hidden.

Plan in plain language

The plan is a markdown draft: what changes, what tests, what risks. You read it, you push back, you edit. Nothing runs until you approve.

plan.draft — no hidden steps.

Execute in a worktree

Approved cards run inside an isolated git worktree on a dedicated branch. Tests run first, diff is staged, main is never touched without a human press.

git worktree, ephemeral.

Ship a diff, not a surprise

You review the diff in your editor of choice. If it's wrong, you throw the worktree away and the card goes back to Planning. If it's right, you merge.

No auto-merge. Ever.

Ideation

3 cards

Backlog

11 cards

Bugs

2 cards

In Progress

1 card

Test

1 card

Completed

47 cards

§ 05from the authorwhy this exists

I spent twenty years building enterprise software, and the whole time, the ability to hold a complex chain of thought from “rough idea” to “shipped system” was the exclusive property of a small group of specialists — architects, lead engineers, senior analysts. Everyone else was structurally locked out.

Then AI changed the unit economics. A founder with taste can prototype in a weekend. A domain expert can describe a workflow and have working code thirty minutes later. That much is real.

But the shift I care about isn't “everyone uses AI.” It's everyone collaborates with AI as a team — and inherits an enterprise-grade chain of thought capability they never had access to before.

Ideafy is my attempt to bring people one step closer to that. Not AI as an assistant on the side. AI as a teammate inside a disciplined surface — columns, plans, tests, verifications — where the kind of thinking that used to belong only to enterprise architects becomes something a founder, an analyst, or a first-time builder can inhabit.

— Ozan

§ 06two editionsship.config

Two editions of the same kanban. Same file, different reach.

Solo is a file on your disk — open source, self-hosted, free forever. Team is the same file shared across a pool, hosted for you, with roles and a claim/push protocol. Pick the one that matches your current shape and switch when it stops fitting.

edition 01 / self-hostedfree forever · MIT

Solo Builder

A single SQLite file, a desktop app, and your CLI of choice. Works offline, owes nothing to anyone.

SQLite database — one file, rsync-friendly backup
Runs fully offline, zero external dependencies
Desktop app for macOS with native window chrome
Bring your agent: Claude Code, Gemini CLI, Codex

clone from github →

edition 02 / managed$5/user/mo · $50/user/yr

Cloud Team

Everything in Solo, plus a shared Pool, role-based access, and managed hosting. The file becomes a place your teammates can meet.

Every Solo feature, without the self-host overhead
Cloud Pool — push a card, pull it back, claim it as yours
File tasks remotely — open a card from the web, your phone, or a teammate's laptop
Per-card discussion — thread feedback into Detail, Opinion, Solution, or Tests
Roles: owner, admin, member — scoped to the workspace
Managed hosting, zero setup, two months free annually

start a cloud team →

01 MIT license02 zero config03 sqlite — no db setup04 works offline

One command to start. Zero excuses left.

Start building in 60 seconds. Keep the receipts forever.

download for macOS start a cloud team →

Apple Silicon (arm64) & Intel (x64) — two DMGs · auto-updates via GitHub releases · cloud hosted in EU

developer? git clone & npm run electron →

// editionsmanifesto·for founders·for developers·docs