SG StoryQuest - Thinking
Version 5 · Last Updated 2025 Oct 21
It has been updated as my thinking changed. See the Changelog at the bottom for what shifted and when.
Core Idea
SG StoryQuest is a sandbox project for testing Agentic Project Management (APM) in practice. It evolved from a simple card game clone into a linear narrative experience — and then into a potential teaching tool. The product is secondary to what the build process revealed.
State of Thinking
The project is shipped, functional, and deliberately paused.
Phase 1 — build and learn — is done. The goal was never to validate a product. It was to test whether APM holds up under real implementation pressure. It does, with one significant gap: context management across sessions is a structural friction the workflow does not yet solve.
The most important shift in this phase was the user reframe. The original assumption was casual players discovering the game organically. That was unrealistic. No distribution mechanism exists. The user who would actually encounter this in a structured way is an educator — and that changes what the product needs to be. That insight belongs to Phase 2.
Thinking Trace
The starting point was not storytelling or games. It was:
I want to understand how to actually use GenAI for development without losing control of what I’m building.
I needed a project — something simple enough to complete, structured enough to expose where APM breaks down. Timeline (the card game) felt right: clear mechanics, minimal ambiguity, easy to scope.
Early on the thinking was almost purely implementation-driven. Pick something, build it with GenAI, see what happens.
The first real friction was gameplay. Timeline felt too binary — right or wrong, nothing in between. Not interesting enough to play, even if easy to build. That discomfort pushed toward a quest-based structure: progression over correctness, some narrative framing.
Then scope started pulling. Branching paths, richer mechanics, open-ended exploration. I cut all of it deliberately. The project existed to test APM, not to build the most interesting game. Linear only. Ship it.
The user assumption broke the moment I actually thought about it seriously. Who discovers this organically? Nobody. There’s no distribution. The realistic user is someone brought to it in a structured context — a classroom, a guided session. That means the educator is the real user, not the student.
That was the biggest shift. But I didn’t chase it. Instead I defined done — one game pack, minimal UI, deployed — and stopped. Phase 1 complete.
What changed?
v5
Shifted from open-ended building toward a defined endpoint — project closed as a learning artifact with Phase 2 explicitly separated.
v4
User assumption flipped — educator reframed as the primary user, casual player discarded as unrealistic.
v3
Scope deliberately constrained to linear design — branching paths and richer mechanics cut entirely.
v2
Timeline mechanics abandoned in favour of quest-based structure — progression and narrative framing replaced binary right/wrong sequencing.
Why it changed?
v5
Mixing “build to learn” with “build to validate” in the same phase creates drift. Separating them forces clarity on what success means at each stage.
v4
Testing the user assumption against reality — even just mentally — made the original assumption collapse immediately. No distribution mechanism exists for casual discovery.
v3
The project existed to test APM, not to build the most interesting game. Complexity was pulling scope in a direction that would obscure the actual learning goal.
v2
Timeline felt too rigid to be worth playing even if easy to build. The discomfort with binary correctness pushed toward something with more room for exploration.
Evidence / Signals
- Product direction shifted four times during implementation — the idea only stabilised under the pressure of actually building
- The educator reframe only emerged when the user assumption was tested against reality, even just mentally — it wasn’t visible during ideation
- Shipping created closure that continued building could not
- Kahoot demonstrates that playful, game-based tools can achieve real adoption in educational settings — the format is not a barrier to educator uptake
- Kahoot’s model is 1-to-many and participatory — which maps well to classroom use, and suggests that format may matter as much as content
Open Questions
- Will educators actually adopt this format without significant hand-holding?
- Is the value in the content, the structure, or the delivery format?
- How much customisation does an educator actually need versus pre-built packs?
- Should Phase 2 treat this as a tool, a content system, or a platform — and does that decision need to be made before validating?
- Kahoot is well-developed and well-received in this space — what does differentiation actually look like against a tool that already works?
Connections
- GenAI Development - From Vibe Coding to APM — underlying method
Changelog
- v5 — MVP shipped; phase separation established; project deliberately paused
- v4 — Educator reframed as primary user; player assumption discarded
- v3 — Scope constrained to linear design; complexity deliberately cut
- v2 — Quest structure introduced; Timeline mechanics abandoned
- v1 — Initial node; Timeline as base; APM test begins