Designed for governed research workflows
The product architecture anticipates authenticated corpora, reviewed Evidence Cards, model orchestration, durable revision history, and human governance. None of those production services are connected here.
A real interaction model with deliberately unreal records.
The showcase demonstrates how MagicForge can connect routine analysis, provenance, revision lineage, and rehearsal observations without connecting to private infrastructure, copyrighted research material, or a live AI Provider.
Method-aware synthetic vertical slice
Three routines, nine synthetic Evidence Cards, nine knowledge concepts, synchronized audience and method timelines, deterministic diagnostics, a local Revision 03, and explicitly unverified observations.
Original, bounded, and traceable
Common technique-family names, project-authored fictional structures, high-level dependencies, angles, timing, failure, recovery, cleanup, reset, and original explanatory wording.
Private or copied instruction
No copied book, course, or video instruction; no protected script; no proprietary authorization claim; no guessed handling presented as fact; no private records, Provider call, account, or database write.
Deterministic does not mean validated
The analyzer is inspectable by design.
Eight explicit phases map declared effect, method, and performance inputs to bounded diagnostics. A rule cannot certify concealment, audience response, or universal method reliability.
- Effect inputs
- Public script, reveal timing, instruction count, confirmation beat, and final-image duration.
- Method inputs
- Family, difficulty, control, force, switch/load, reset, angle, skill readiness, and rehearsal count.
- Performance inputs
- Venue, posture and sight lines, audience size, lighting, and spectator handling.
- Outputs
- Audience, method, performance, risk, and evidence-gap diagnostics with technique, consequences, alternatives, provenance, and review status.
- Revision contract
- Revision 02 remains immutable. Revision 03 records effect, method, performance, risk, evidence, dependency, tradeoff, and required rehearsal test.
- Observation contract
- Technical, performance, and environmental observations update separate hypotheses. Every observation stays synthetic = true, verified = false, and not universally validated.
Public data allowlist
28 explicit records.
The public evidence bundle carries only selected project-owned records. Interactive routine fixtures are defined separately in the client bundle and obey the same synthetic, self-authored, redistribution-allowed boundary.