The Thinking Box

The Thinking Box is the current educational focus of the earlier Human-Led Agentic Product Studio. Its intended audience centres Irish Transition Year learners who may be new to product development, with a facilitator available and professional roles introduced through the work.

The intended outputs are a controlled product demo and a learner-owned presentation explaining the idea, evidence, choices and remaining uncertainty. Meaningful choices should change behaviour, information or rules. A polished result alone cannot demonstrate that the learner understands it.

Programme duration, group ownership, facilitator responsibilities and assessment remain open. The current focus is a coherent learning experience before beta, rather than a promised school programme.

The learner gives an idea context, investigates one question and records what they saw, read or heard. Watch, Try, Read and Ask activities provide a way in without requiring AI. A separately saved interpretation explains what the notes might mean and what remains unknown.

Continue, Revise, Park and Stop are deliberate human choices, each with a reason. A current interpretation and recorded evidence are required before a decision. These checks establish completion of the workflow; they do not prove demand, evidence quality or learning.

After Continue, the learner develops the intended audience, need and proposed experience. Saved context, interpretation and decision carry into a readable project brief with contents, numbered sections and the complete original evidence and sources.

The document preserves the learner's words and paragraph breaks. Previewing, submitting and reviewing use the same structure. Each submission is a fixed version: later edits cannot rewrite the evidence an earlier review was based on.

Admin can request changes or approve progression, with feedback required for either. Approval records a review outcome; it does not unlock a Model stage that has not been built.

Two optional AI paths are connected. A Product Manager explains concepts, suggests activities and drafts idea notes when asked. A Product Researcher challenges a saved interpretation with possible overstatements, alternatives and questions tied to the supplied evidence.

Sending a request and accepting proposed idea notes are explicit actions. Accepted notes remain labelled AI-assisted planning material; they do not become evidence, the learner's interpretation, a decision or Admin approval. Other professional roles remain planned.

The Product Manager receives saved project context, but structured brief fields and Admin feedback are not yet included. That missing connection is a defined next increment, not a capability being claimed today.

The local application combines a React and TypeScript interface, a Node service and SQLite records. Revision checks protect against silently overwriting newer work; project histories, draft recovery, Markdown reports and validated backup-and-restore support continuity.

AI credentials stay on the server, while Product Manager conversation state also persists at the provider. This is not a zero-retention system. The build uses fictional, non-sensitive practice material and a self-selected User/Admin rehearsal switch.

Hosted learner use still requires real authentication, protected Admin assignment, project-level access controls, appropriate data handling and operational recovery. The present implementation remains local.

The intended sequence is Think, Model, Prototype, Learn, Decide, Architect, Build and Protect. Only Think is implemented. Its question-level decision is distinct from the later Decide stage; the demo and presentation are not yet available end to end.

The next work is to close the current brief and review loop, then carry one School Gallery journey into Model and a bounded prototype. School Swap and School Mixtapes are reference concepts for later challenges, not finished applications.

Founder testing and scoped technical checks provide useful development evidence. Whether learners understand the work, make better decisions and can explain a product they have helped create still needs independent observation.