| Milestone | Term | Target date | Status | |
|---|---|---|---|---|
| 1 | Project scoping and design (FYP-1): architecture decisions, communication protocol, sandbox design, dashboard model, form alignment, supervisor confirmation on evaluation scope | Term 1 | Completed 27/09/2026 | Done |
| 2 | Core system implementation (FYP-2): data models and ticket FSM, config and loader, orchestrator loop, sandbox controller, Docker access | Term 1 | Completed 25/09/2026 | Done |
| 3 | Dashboard and UI (FYP-3): chat feed and kanban board, custom roles editor, intervention controls | Term 1 | Completed 25/09/2026 | Done |
| 4 | Evaluation harness (FYP-4): task and scoring models, hidden test isolation, MAST failure tagger, single-agent baseline, runner and CLI | Term 1 | Completed 25/09/2026 | Done |
| 5 | Real Claude-backed agent across every ticket stage, and the first real end-to-end pipeline run on the live API | Term 1 | Completed 28/09/2026 | Done |
| 6 | Option B: hierarchical-versus-flat comparison runner and LLM rubric judge, run live against real tasks | Term 1 | Completed 29/09/2026 | Done |
| 7 | Human-eval study delivery mechanism (FYP-36) | Term 1 | Completed 29/09/2026 | Done |
| 8 | Dashboard visual redesign and interactive kanban (FYP-34, FYP-35) | Term 1 | Built 29/09/2026 | Done, pushed to its own branch, not yet merged to master |
| 9 | Top-down 3D office visualization: static shell (FYP-37), agent characters at their home stations (FYP-38), navigation with the walk cycle (FYP-39), the behaviour model with its ticket failure paths (FYP-40), the scenario library with playback controls (FYP-41) and observation with the event log (FYP-42) all done. Only the live orchestrator event stream (FYP-43) remains, and it was deliberately deferred | Term 1 | Layers 1 to 6 done 30/09/2026 and 01/10/2026; Layer 7 deferred | Built, Layers 1-6 of 7; Layer 7 deferred by design |
| 10 | Project poster (FYP-45 to FYP-48 done): example posters studied, content chosen, built as a page, diagrams and screenshots prepared, symmetry measured, exported as a PDF. The human evaluation study has a reserved space. Ethics approval, the poster's requirements and the final pre-print checks are open (FYP-49 to FYP-52) | Term 1 | Draft done 05/10/2026 | Draft built; not for printing or submission yet |
| 11 | First supervisor meeting with Ziyang | Term 1 | 08/10/2026, 3:30pm to 4pm | Scheduled |
| 12 | Documentation and report (FYP-5): the dissertation itself, chapter by chapter | Term 1 | TBC | Not started |
Target dates marked TBC are genuinely not set yet. I will fill them in once I have real deadlines from the module handbook or from Ziyang, rather than guess at them here.