What do you want your intelligence to become?
Anyone can shape intelligence. You don’t need to know how to train a neural network.
Shape is the creative environment of modelparadox, the platform for shaping intelligence. You develop an AI model from your own words, examples and judgements; the engineering underneath is being built.
What you shape is a Creation.
In shaping, an intention becomes a Creation: you keep developing its model, one version after another. Not a chat that ends, a prompt or a saved setting.
An illustrative drawing follows. The working example is in the next section.
- intention:
“I want to develop a model that teaches mathematics differently. It should help people discover their mistakes rather than immediately giving them answers.”
- v1: Starts from an existing model. It gives answers straight away.
- Explains clearly, checked
- Lets people find the mistake, not yet checked
- possible behaviours: None of them yours yet.
- demonstration: Your worked examples, with the mistakes left in.
- distinction: A hint is not the answer.
- v2: An adaptation the collaborator proposed, learned from your examples.
- Asks what they tried, checked
- Stuck learners, not yet checked
- experiment: Show the answer after two tries.
- failure: They stopped trying and waited.
- experiment: Point to the step, not the answer.
- v3, refinement: Points to the step that went wrong, never the answer.
- Finds the wrong step, checked
- Slower to help, worse
- Showing the answer, not kept
- the model underneath changes
- v4: Moves to a different model underneath. The adaptation is learned again. Same Creation. Every check runs again.
- Asks what they tried, checked again
- Finds the wrong step, checked again
- Slower to help, still worse
- Stuck learners, not yet checked
- still being shaped
Shape is where you develop it.
Begin with what you want, in your own words. A collaborator proposes; every decision stays yours. In this prototype it is scripted for the example.
- Develop a modelStart from your intention. No model to pick and nothing to set up first.
- Bring examples and knowledgeAdd examples, demonstrations and notes, each with a clear role.
- Investigate alternativesWrite a situation, take over a response and continue it.
- Understand changesProposals, decisions and versions are recorded, and their details are one step away.
- ContinueYour work is saved in this browser. Come back and carry on.
A worked example: a model that helps someone understand their own mistakes, with demo history and simulated results, each labelled.
Open the example in ShapeWorking prototype, 17 October 2026. It runs in your browser without an account. Its collaborator is a scripted demonstration, and training is simulated.
A programmer works with instructions. A designer works with visual and interactive elements. A shaper works with intelligence itself.
Most AI tools let you use a model, or configure one with prompts. Shaping is developing the model itself: what it can do, how it behaves, what it knows and what it must never do.
The idea is that you work with an AI collaborator: you show it examples and data, look at the changes it proposes to the model, run experiments and see what changed, version by version. The interface stays simple. The engineering underneath does not.
We aren’t building intelligence for everyone. We’re building a world where everyone can participate in creating it.
| Medium | Medium · Human activity | What people create |
|---|---|---|
| Language | LanguageWriting | Expressed ideas and knowledge |
| Computation | ComputationProgramming | Executable software |
| Visual design | Visual designDesigning | Visual and interactive experiences |
| 3D environments | 3D environmentsModelling | Digital objects and worlds |
| Machine intelligence | Machine intelligenceSHAPING | Intelligent capabilities and behaviours |
Intention guides the work. Evidence establishes what was achieved.
Shaping asks two questions: is this becoming what I had in mind, and did the change work? The first is yours to answer. The second takes serious engineering.
Today, developing a model means stitching together many separate tools. modelparadox is building one foundation for that work, so the engineering can stay rigorous while shaping stays simple.
v4, on a different model underneath
The adaptation was learned again on the new model. This is what was checked again, and what wasn’t.
- Asks what they triedChecked again
- Finds the wrong stepChecked again
- Speed of helpSlower than v2
- Stuck learnersNot yet checked
- Covers v4 only.
- Says what was checked, not that the Creation is good.
- EngineBeing built to run and record experiments, training and model adaptations.in development · runs locally
- DataThe material behind a Creation, and where it came from.in design
- EvaluationWhether a change did what was meant, and what it broke.in development
- EvidenceWhat ran, what changed, and what was or wasn’t checked.in development
- RuntimeRunning a Creation where it’s needed.planned
- CloudHosted workspaces and deployments.planned
- BusConnections to models, tools and the outside world.in design
- ProtocolShared definitions every part works to.in design
Statuses as of 14 October 2026. None of this is publicly available yet, and the Shape prototype does not use it: its engineering is simulated.
Shaping is bigger than one platform.
modelparadox sits within a wider ecosystem. The idea connects an intention, a model you develop, evidence of what it does and a model others can reuse, and it may eventually lead to distribution and potential income for the people who made it. None of it is a step you have to take.
- openparadoxOpen research, models, tools, experimentation and accessible learning. Nothing has a public release yet.openparadox.ai
- MarketplaceMay eventually let people discover, use and distribute the models they develop, and perhaps earn from them.planned
- el-xMay assist with shaping.not yet public
Statuses as of 14 October 2026.
Don’t program intelligence. Shape it.
Create it. Shape it. Understand it. Run it. Share it.
That is the experience we’re building towards. Shape is where it begins.