AI-IPBL in real classrooms
Working with teachers to bring the nine-stage workflow into classrooms, observing how an AI Agent supports sustained student inquiry.
AI for Human Creativity & Learning
What happens when AI doesn't just answer questions, but stays with a project?
FeiyiAI is building project-centered AI: an AI Project Agent that remembers every discussion, decision and reflection, working alongside learners and creators to turn a question into a real body of work.
feiyi.ai/workshop
Why Feiyi
AI can answer questions. But can it stay with a project?
Conversational AI is good at giving answers. Learning and creation need lasting context, judgment and companionship. Here is the difference:
| Typical AI assistant | Feiyi Workshop | |
|---|---|---|
| Project memory | Context is forgotten when the chat ends | Project Memory keeps every discussion, decision and reflection |
| Participation | One question, one answer, full stop | Stays through all nine stages of the workflow |
| Relationship | Answers for you | Asks back and reflects, so answers grow from your own inquiry |
| Outcome | A block of text | Real artifacts: research notes, prototypes and works |
Workflow
From question to outcome.
Nine stages, one complete project path. The AI Project Agent supports each stage with full project context.
- 01
Research
Build project context
- 02
Question
Sharpen the question
- 03
Concept
Diverge on ideas
- 04
Experiment
Run small tests
- 05
Prototype
Draft it hands-on
- 06
Critique
Trace the evidence
- 07
Production
Polish the artifact
- 08
Exhibition
Present the work
- 09
Portfolio
Archive everything
Features
One workspace for the whole project.
01
Nine-stage workflow
A clear track from inquiry to reflection. The AI knows where you are and what comes next.
workflow · 9 stages
Inquiry
Research
Concept
Experiment
Prototype
Critique
Outcome
Showcase
Reflection
02
Multi-agent collaboration
Specialized agents for analysis, suggestion and reflection relay work on the same project memory — a small team around your project.
agents · relay
agent-1
Analyze
Reads project context
agent-2
Suggest
Proposes next steps
agent-3
Reflect
Traces decisions
03
Project Memory
Every discussion, decision and revision is remembered. The agent cites your project history, not guesses.
project-memory
How do I push them to dig deeper?
Revisit the three stage-2 discussions and ask from there.
memory
discussion#12
decision#3
revision#7
04
Artifact archive
Outlines, prototypes, reports and presentations grow into an archive as the project moves.
artifacts
Research outline v3
Usability test report
Presentation deck
9
workflow stages
100%
open methodology
zh·en
bilingual by design
Projects
The first projects are on their way.
An AI workflow for creators
Inviting independent creators to complete a work on top of Project Memory, recording every decision and reflection along the way.
Outcomes of the first projects will be published with research notes.
Research
We publish the process, not just conclusions.
Research notes, method retrospectives and project reports, made public as work progresses.
AI should amplify human thinking, not replace it.
Follow what we are building, or talk with us about your project or classroom.