For people about to build with AI
Your idea isn't clear yet.The AI is already coding.
GoodIdea has AI weigh in from research, design and engineering first.The trade-offs that matter are yours to decide — then it becomes a first version a coding agent can build.
About 76 seconds · a public demo; the workbench isn't open yet
Designer quoting tool
The first idea
Could I have a tool that makes quoting faster?
Your call
In version one, can the client adjust the scope?
Once you decide: v0.2, then a first-version scope and a handoff
The demo
Watch one quoting-tool idea get specific.
The case: a designer wants a quoting tool for herself. Three AI roles weigh in, and partway through there is one trade-off you decide for her.
Output
The same case. This is what is left.
When the film above finishes, this is what remains: a product draft, the scope of version one, and the package a coding agent would be handed.
The case: a freelance designer who wants the scope they re-explain on every quote to become one sheet a client can read and confirm.
How it works
The agents propose. GoodIdea combines. You decide.
Three steps, then three principles — and the principles all answer one question: when it gets something wrong, can you pull the conclusion back?
- 01
Specialists propose, from different angles
Research reads what the client keeps asking. UX sketches where the edges are. Engineering prices the build. Each contribution is signed, and says whether it is an assumption, an exploration or a suggestion.
- 02
GoodIdea states the direction and the trade-off
The three contributions come together into what they agree on, plus the one choice that needs a person to make it. Disagreement is put on the table, not smoothed away.
- 03
Nothing reaches the draft until you confirm it
The proposal waits. After you confirm, the old description is struck out, the new direction is written in and the version moves — so you can see what your decision changed.
- Contributions are attributed
- Every line on the draft says which agent's contribution it came from.
- Proposed and adopted stay apart
- An unconfirmed proposal keeps its own dashed edge; it never quietly becomes product content.
- What is unverified is marked
- Research proposes an assumption to test; a sketch is an exploration, not an adopted design. Both say so.
Editing on the map itself, and asking for changes inside a running prototype, are where GoodIdea is going. The public film does not have those interactions yet.
About this project
GoodIdea is a project still being worked out.
The question behind it: when AI can keep proposing, how does a person stay in charge of where the product goes and of the final call? The 76-second demo on this page is one answer so far.
What is public today is this one fixed case and the interface it runs on. The live workspace is not open yet: nothing on this page talks to a backend, nothing is really generated, and no package is really sent to a coding agent.
Watch the demo again