Invite-only access

AI made building cheap. Code, prototypes, mockups, MVPs — they can be generated faster than ever.

Knowing what to build
is the new bottleneck.

ForkIdeas helps founders, teams, and AI agents decide what is worth building before they waste time shipping the wrong thing. Structure the thesis. Expose weak assumptions. Fork alternatives. Validate direction. Move to GitHub only when the idea is build-ready.

Chat is ephemeral. Decisions are not.

$ fork init "AI-powered onboarding for SaaS"

✓ Repo created — v0.1.0 (decision state)

$ fork critique --depth=deep

⚡ 3 weak assumptions found, 2 untested risks flagged

$ fork branch --reason="alt direction"

⑂ Fork created — exploring a second direction

$ fork commit

→ Committed v0.2.0 — decision updated, evidence linked

The real problem

Teams build faster than they decide.

AI amplified execution. But most teams still run on invisible decisions. They move fast, but they don’t track why the direction changed. They don’t remember what failed. They repeat the same mistakes in new forms.

Ideas are generated fast — but selection criteria stay fuzzy.

MVPs ship quickly — but assumptions are rarely explicit.

Iteration happens — but learning is not recorded.

There is no decision history you can trust.

Risks and assumptions hide inside docs and chats.

The same mistakes reappear because the system forgets.

If you can’t see the decision trail, you can’t improve it.

Positioning

ForkIdeas is a decision system for what gets built.

Software development is no longer expensive. Decision quality is. ForkIdeas turns messy debate into structured, reviewable decision state.

ForkIdeas is NOT

Idea management.

Brainstorming.

Note-taking.

ForkIdeas IS

A system to decide direction before code.

A place to validate assumptions with evidence.

A way to track what changed and why.

A faster way to kill weak directions and commit to strong ones.

Decision workspace

Not chat. Not comments. A decision workspace.

ForkIdeas stores decisions as objects: thesis, assumptions, evidence, critique, forks, and validation. Nothing important stays trapped inside chat threads. Everything is inspectable, challengeable, and build-ready when it survives pressure.

onboarding-saasv0.2.0
thesis

statement: "Onboarding completion increases retention for PLG SaaS"

icp: "PLG SaaS, 10-50 seats"

scope: "MVP: onboarding checklist + alerts"

assumptions

A1: "Personalized onboarding increases activation by >20%"

A2: "Teams will pay for onboarding automation"

evidence

E1: "A/B test confirms onboarding completion -> retention"

E2: "12 interviews show onboarding is a top-3 churn driver"

forks

F1: "Fork: SMB self-serve workflow"

F2: "Fork: enterprise onboarding teams"

Decision tracking

Every decision is recorded. Every change has a reason.

If you can't point to when your thinking changed, you didn't actually learn.

ForkIdeas keeps a versioned trail of decisions. Evidence links to the exact change it caused.

v0.1.0seed
Initial direction

A starting thesis with explicit assumptions. No code yet.

v0.2.0challenged
Assumption invalidated

A critique surfaced weak assumptions. Direction narrowed. New validation plan added.

~ Scope narrowed Assumption A2 invalidated+ 3 evidence items added
v0.3.0validated
Build-ready direction

Core thesis holds under pressure. MVP scope locked with traceable reasoning.

+ Validation outcome recorded+ MVP spec generated~ Decision updated
Artifacts persist as part of the decision trail:
critique_report
Structured critique with severity
risk_analysis
Risks with mitigation paths
validation_plan
Experiments mapped to assumptions
mvp_spec
Scoped MVP with priorities
Forking

Forking is decision exploration.

Instead of arguing what is right, you explore multiple directions in parallel.

Fork different ICPs, pricing, workflows, or product shapes. Let evidence decide what deserves execution.

Onboarding for PLG SaaSv0.2.0 — origin
Fork 1:Enterprise motionevolving

Same thesis, different buyers and workflow

Fork 2:Self-serve templatesinvalidated

Reduce complexity and test a simpler product shape

Fork 3:Platform anglevalidated

Different direction: product becomes a platform

Parallel exploration

Explore multiple directions without premature consensus.

Evidence decides

Compare forks against outcomes and make a decision.

Traceable reasoning

Forks keep history — you can see why one won.

Human + AI

This is not chat. This is shared decision state.

AI changed building. It did not solve judgment. Humans decide what matters. AI structures thinking, finds gaps, challenges assumptions, and remembers everything. Both operate on the same persistent system state — not comments and not ephemeral threads.

Humans
Judgment
Decide priorities, interpret evidence, pick the direction.
Validation
Run experiments, talk to users, bring evidence back.
Conviction
Commit to a build-ready MVP when the reasoning holds.
AI
Structure
Turn messy input into explicit decision objects.
Challenge
Find gaps, weak assumptions, and untested risks.
Memory
Persist critiques, outcomes, and decision history.
Re-entry

When you come back to something you were building, you see the decision state.

You immediately know what changed, what failed, what was validated, and what needs action. No re-reading. No archaeology.

Snapshot (build context)

What changed since last time

What was validated vs invalidated

What is build-ready vs blocked

What needs action next

Decision history

2026-04-10 validated ICP narrowed: PLG SaaS (10-50 seats)

2026-04-11 warning Risk flagged: willingness-to-pay not tested

2026-04-12 validated Retention link validated via experiment

Full context in seconds — not minutes.

Multilingual system

Language is a view layer. Meaning is the system.

Switch the UI language. Contribute in any language. ForkIdeas stores meaning and structure underneath, so teams and AI can collaborate across language boundaries without losing intent.

UI language switcher (header)
Outcome

You stop building things that should never be built.

Building is cheap. Being wrong is expensive. ForkIdeas makes wrong directions visible early — and makes strong direction easier to commit to.

Validated before built

Evidence and critique happen before code and sunk cost.

Explored before committed

Multiple directions evolve in parallel until one wins.

Faster to kill, faster to commit

Weak directions die earlier; strong ones gain conviction faster.

AI that compounds

Artifacts persist. Context improves. Decisions get sharper over time.

Invite-only. By design.

ForkIdeas is invite-only. We optimize for depth of engagement, not volume. Users earn the ability to invite others through contribution quality.

01

Request access or receive a code from an existing member

02

Enter your invite code to unlock your account

03

Contribute. Earn invites. Bring in people who think rigorously.

Don't start with code. Start with decision quality.

Make assumptions explicit. Track what changes. Fork directions. Validate what matters. Then build with confidence.