docs: reconcile consolidated defense story

This commit is contained in:
lda
2026-07-13 14:37:56 +07:00 Verified
parent c3bef1dc8a
commit 7a4554b21e
21 changed files with 209 additions and 205 deletions
+10 -10
View File
@@ -148,8 +148,8 @@ location.reload();
When a stateful demo deep link is opened before this browser session has started
a live run, the presentation selects the reviewed recording for that beat even
if the loopback health probe succeeds. Scene 8's request and Send action remain
local replay behavior; start the live path from the Scene 10 run operation.
if the loopback health probe succeeds. Scene 7's request and Send action remain
local replay behavior; start the live path from the Scene 9 run operation.
Otherwise the typed interrupt, output, and trace links remain useful
replay-backed entry points.
@@ -256,19 +256,19 @@ Say:
## Twenty-Five-Minute Timing
Target an 11:45 must-say path plus 1:15 for navigation and demo transitions,
Target an 11:00 must-say path plus 1:15 for navigation and demo transitions,
leaving 12 minutes for questions.
| Time | Segment | Notes |
|---:|---|---|
| 0:00-1:30 | Title and problem | AI-agent goal becomes a workflow substrate contribution |
| 1:30-2:15 | Positioning | Adjacent systems and the prototype's narrower position |
| 2:15-5:30 | Architecture and authoring | Planner/runtime boundary, lifecycle, implementation, authoring |
| 5:30-8:30 | Prepared demonstration | Request, deployment, start, interrupt, resume, output, trace |
| 8:30-10:30 | Evaluation | 36 audited trials as bounded engineering evidence |
| 10:30-11:45 | Limits and conclusion | Security, scheduling, evaluation, and agent-layer boundaries |
| 11:45-13:00 | Transition buffer | Navigation, interaction, and demo delay |
| 13:00-25:00 | Questions | Use prepared Q&A branches when available |
| 2:15-4:36 | Architecture and lifecycle | Planner/runtime boundary, lifecycle, and architecture |
| 4:36-7:45 | Prepared demonstration | Request, authoring, deployment, start, interrupt, resume, output, trace |
| 7:45-9:45 | Evaluation | 36 audited trials as bounded engineering evidence |
| 9:45-11:00 | Limits and conclusion | Security, scheduling, evaluation, and agent-layer boundaries |
| 11:00-12:15 | Transition buffer | Navigation, interaction, and demo delay |
| 12:15-24:15 | Questions | Use prepared Q&A branches when available |
Use [`defense-speech-and-claim-audit.md`](defense-speech-and-claim-audit.md)
for the rehearsed speech, evidence qualifications, and prioritized Q&A.
@@ -316,7 +316,7 @@ Answer:
## Pre-Defense Checklist
1. Start `/present` with the intended target. Verify Scene 8 submits the prepared
1. Start `/present` with the intended target. Verify Scene 7 submits the prepared
request, and verify the prepared-run action appears on the demo scenes when
the live badge is ready.
2. Complete one submitted branch and confirm the same run reaches output and
+32 -44
View File
@@ -1,6 +1,6 @@
# Defense speech and claim audit
This runbook gives you the must-say 11:45 defense speech, a 1:15 navigation buffer, and a prioritized 12-minute question period. It also marks claims that require qualification so the presentation stays aligned with the thesis and current implementation. The typed catalog at `web/apps/console/src/presentation/presenter/presenter-notes.ts` is the source for the must-say text.
This runbook gives you the must-say 11:00 defense speech, a 1:15 navigation buffer, and a prioritized 12-minute question period. It also marks claims that require qualification so the presentation stays aligned with the thesis and current implementation. The typed catalog at `web/apps/console/src/presentation/presenter/presenter-notes.ts` is the source for the must-say text.
## Timing and evidence rules
@@ -11,18 +11,18 @@ Use these evidence labels while rehearsing. Do not read the labels aloud.
- **Qualify**: accurate only within an explicit boundary
- **Do not claim**: unsupported, untested, or excluded from scope
Target 11:45 for the must-say speech. Keep 1:15 for navigation or demo delay; the complete deck target is 13:00. The question period then has 12 minutes.
Target 11:00 for the must-say speech. Keep 1:15 for navigation or demo delay; the complete deck target is 12:15. The question period then has 12 minutes.
| Segment | Target |
| --- | ---: |
| Scenes 1-2: goal and problem | 1:30 |
| Scenes 3-7: positioning, model, and implementation | 3:15 |
| Scenes 8-12: prepared demonstration | 3:00 |
| Scene 13: evaluation | 2:00 |
| Scene 14: limits and conclusion | 1:15 |
| Must-say speech | 11:45 |
| Scenes 3-6: positioning, model, and architecture | 3:06 |
| Scenes 7-11: prepared demonstration | 3:09 |
| Scene 12: evaluation | 2:00 |
| Scene 13: limits and conclusion | 1:15 |
| Must-say speech | 11:00 |
| Navigation buffer | 1:15 |
| Complete deck target | 13:00 |
| Complete deck target | 12:15 |
## Main speech
@@ -92,32 +92,20 @@ Keep Scene 5 conceptual. Scene 9 applies this vocabulary to the prepared example
### Scene 6: Zoom through the implemented architecture
**Route:** `architecture/overview`, `architecture/client`, `architecture/api`, `architecture/runtime`, then `architecture/node-use`<br>
**Time:** 3:55-4:50
**Route:** `architecture/overview`, `architecture/client`, `architecture/api`, then `architecture/runtime`<br>
**Time:** 3:55-4:36
**Evidence:** Supported
Say:
> First, the implemented architecture spine and its ownership boundaries. Humans and agents share one public lifecycle surface. WorkflowApi owns lifecycle operations; JSON-RPC only adapts transport. WorkflowServer composes records, capabilities, API, and kernel; providers remain outside the core. NodeUse invokes a NodeDef handler, reduces state, records trace, and routes the declared outcome.
> First, the implemented architecture spine and its ownership boundaries. Humans and agents share one public lifecycle surface. WorkflowApi owns lifecycle operations; JSON-RPC only adapts transport. WorkflowServer composes records, capabilities, API, and kernel; providers remain outside the core.
The NodeUse sequence is documented in the thesis runtime diagram and narrative ([lines 706-750](../thesis/system-design-implementation.md#workflow-core-model)).
The optional NodeUse deep dive remains available through the architecture focus route and Q&A; it is not part of the timed forward sequence.
### Scene 7: Explain agent-operable authoring
**Route:** `authoring/discover`, `authoring/author`, `authoring/diagnose`, then `authoring/repair`
**Time:** 4:50-5:30
**Evidence:** Supported, with qualification
Say:
> Before authoring, a client can discover sources, capabilities, and schemas instead of guessing at hidden interfaces. Focused operations let an external agent change a mutable Draft while preserving a clear lifecycle boundary. Validation returns structured diagnostics, affected paths, repair hints, and suggested next actions. These surfaces make invalid intermediate drafts repairable; they support a loop, but no hint guarantees success.
Do not say that every repair hint guarantees a valid workflow. Say that diagnostics are designed to support repair loops and that targeted tests cover the implemented paths ([lines 909-962](../thesis/system-design-implementation.md#validation-and-diagnostics)). Mention the instruction layer if asked: skills and runbooks also affected operability ([lines 1505-1531](../thesis/system-design-implementation.md#agent-instruction-layer)).
### Scene 8: Introduce the prepared demonstration honestly
### Scene 7: Introduce the prepared demonstration honestly
**Route:** `agent-handoff/request`
**Time:** 5:30-5:50
**Time:** 4:36-4:56
**Evidence:** Implementation extension
Say:
@@ -126,22 +114,22 @@ Say:
If replay is active, say: “This is the reviewed recording, not a live model planning this workflow.”
### Scene 9: Show authoring and deployment without starting a run
### Scene 8: Show the prepared lifecycle
**Route:** all five `prepared-lifecycle/*` beats
**Time:** 5:50-6:35
**Evidence:** Implementation extension
**Route:** all six `prepared-lifecycle/*` beats
**Time:** 4:56-5:50
**Evidence:** Implementation extension; supported with qualification
Say:
> The later issue-review example first inspects configured local.lda_docs, report, and issue-board capabilities. It creates and edits a Draft for report generation, making the proposal visible before execution. It validates incomplete state, exposes a missing output binding, and applies a targeted repair. It saves the validated plan as immutable artifact lda_report_case_study version 1. This later issue-review example is richer than the thesis three-node deterministic report case study: it is an implementation extension built on the same platform. Deployment binds and validates a ready configuration but does not run the workflow.
> It inspects sources, capabilities, and schemas rather than guessing at hidden interfaces. Focused operations modify mutable authoring state before execution. Structured diagnostics identify the missing output projection. One focused output-map edit resolves it; hints do not guarantee automatic repair. It saves the validated plan as immutable artifact lda_report_case_study version 1. Deployment binds and validates three local sources; execution starts in the next scene.
This issue-review workflow is repository code under `examples/lda_report_workflow`. It is richer than the thesiss documented three-node deterministic report case study. Call it a later demonstration built on the same platform, not the exact thesis case study.
This later issue-review example is richer than the thesis case study; do not present its issue-board output as thesis output. Diagnostics support a repair loop but do not guarantee automatic repair.
### Scene 10: Start the prepared workflow
### Scene 9: Start the prepared workflow
**Route:** `run-from-deployment/input`, `run-from-deployment/operation`, then `run-from-deployment/graph`
**Time:** 6:35-7:10
**Time:** 5:50-6:25
**Evidence:** Implementation extension; live-capable
Say:
@@ -150,10 +138,10 @@ Say:
If live execution has not been completed during rehearsal, say: “The operation view is replay-backed evidence of the prepared path. I am not presenting this as a newly completed live run.”
### Scene 11: Present a typed interrupt, not a production approval system
### Scene 10: Present a typed interrupt, not a production approval system
**Route:** `typed-human-boundary/interrupt` then `typed-human-boundary/approval`
**Time:** 7:10-7:40
**Time:** 6:25-6:55
**Evidence:** Implementation extension; qualify
Say:
@@ -162,22 +150,22 @@ Say:
Do not call the negative path “deny without resuming.” Both outcomes resume execution through different workflow branches. Do not imply that the prepared revision recording preserves the submitted branchs run identity.
### Scene 12: Show output and inspectable evidence
### Scene 11: Show output and inspectable evidence
**Route:** `resume-output-evidence/resume`, `resume-output-evidence/output`, then `resume-output-evidence/trace`
**Time:** 7:40-8:30
**Time:** 6:55-7:45
**Evidence:** Implementation extension; replay continuity differs by branch
Say:
> On the submitted path, workflow.runs.resume continues the recorded interrupted Run. The workflow creates the report and issue-board changes, then records terminal output. Trace frames and protocol evidence remain inspectable; this is declared-boundary resumability, not arbitrary crash recovery or exactly-once execution. The revision replay is a separate prepared recording.
For the submitted replay, the same run ID is demonstrated. The prepared revision replay currently uses `run_recorded_lda_report_revision`; describe it as a separate prepared branch recording. The thesis itself documents a separate three-node report case study without issue-board mutation, so identify these issue-board results as evidence from the later example implementation.
For the submitted replay, the same run ID is demonstrated. The prepared revision replay currently uses `run_recorded_lda_report_revision`; describe it as a separate prepared branch recording.
### Scene 13: Explain what the evaluation proves
### Scene 12: Explain what the evaluation proves
**Route:** `evaluation/cohort`, `evaluation/validity`, then `evaluation/findings`
**Time:** 8:30-10:30
**Time:** 7:45-9:45
**Evidence:** Supported, with strict qualification
Say:
@@ -186,10 +174,10 @@ Say:
The product and prompts evolved across waves. Call this longitudinal engineering evidence, not a controlled benchmark ([lines 1287-1339](../thesis/system-design-implementation.md#formative-agent-trial-findings)).
### Scene 14: Close on the bounded contribution
### Scene 13: Close on the bounded contribution
**Route:** `conclusion/limits`, `conclusion/future`, `conclusion/conclusion`, then `conclusion/questions`
**Time:** 10:30-11:45
**Time:** 9:45-11:00
**Evidence:** Supported
Say:
@@ -316,5 +304,5 @@ Start with the short answer. Expand only when the examiner continues.
## Presentation defects to resolve or avoid
- The thesiss deterministic report case study has three workflow nodes. The later presentation example has eleven plan nodes, while its simplified graph intentionally omits the terminal `end_cancelled` marker. Avoid quoting a graph-node count unless the distinction is relevant.
- The live health probe does not establish a completed live Scene 10-12 rehearsal. Label replay-backed output and trace evidence as recorded.
- The live health probe does not establish a completed live Scene 9-11 rehearsal. Label replay-backed output and trace evidence as recorded.
- The revision replay has a separate run ID. Never use the submitted branchs “same persisted run” wording for that branch.
@@ -32,9 +32,9 @@ factual or product follow-up rather than silently treating it as solved.
| Scene 1 opening | Product goal, implementation boundary, and planner/tool-surface/runner decomposition are distinct. The opening caption remains editorial and borderless; its row sizes to content instead of clipping the lower padding. |
| Scene 5 to 7 story | Lifecycle vocabulary now leads into authoring and repair rather than presenting authoring as an isolated diagnostic screen. |
| Scenes 7 and 9 | Authoring and prepared-lifecycle evidence use a dominant artifact with secondary assistant/support surfaces. |
| Scene 10 graph | The graph is horizontal, factual, selectable, draggable/zoomable, and fit to the initialized React Flow viewport. Outcome shapes and edge labels remain visible. |
| Scene 13 findings | Six findings use a readable 3x2 layout with icons, campaign size, outcome counts, and the bounded-evidence statement visible together. |
| Scene 14 conclusion | The typed substrate and closing boundary are dominant; non-claims and future-work layers recede without being removed from their own beats. |
| Scene 9 graph | The graph is horizontal, factual, selectable, draggable/zoomable, and fit to the initialized React Flow viewport. Outcome shapes and edge labels remain visible. |
| Scene 12 findings | Six findings use a readable 3x2 layout with icons, campaign size, outcome counts, and the bounded-evidence statement visible together. |
| Scene 13 conclusion | The typed substrate and closing boundary are dominant; non-claims and future-work layers recede without being removed from their own beats. |
## Remaining Findings
@@ -50,7 +50,7 @@ factual or product follow-up rather than silently treating it as solved.
### PRODUCT
- Live Scene 10-12 execution is health-checked and connectable, but still needs
- Live Scene 9-11 execution is health-checked and connectable, but still needs
a complete rehearsed start -> interrupt -> resume -> output -> trace path on
the running workflow server.
- The presentation chat remains a scripted/replay product surface. A live LLM
+6 -6
View File
@@ -21,15 +21,15 @@ rehearsal pass.
|---|---|---|---|---|---|
| PASS | 1280x720 | replay | `/present#scene/agent-handoff/request` | Open the request route and click `Send`. | The prepared conversation appears with the user request, assistant narration, and four discovery tool calls; no run is claimed. |
| PASS | 1280x720, 1024x768 | replay | `/present#scene/typed-human-boundary/approval` | Open the approval route. | The input files, typed interrupt payload, proposed issue, resume comment, `Submit`, and `Request revision` controls are visible. The footer reports `Run paused - review required`. |
| PASS | 1280x720 | replay | `/present#scene/prepared-lifecycle/{discover,draft,validate,artifact,deployment}` | Open each Scene 9 beat directly. | The phase rail advances through Discover, Draft, Validate, Artifact, and Deployment; staged chat groups and factual evidence change with each beat. |
| PASS | 1280x720 | replay | `/present#scene/run-from-deployment/operation` | Open the Scene 10 operation beat. | `workflow.runs.start` is shown as interrupted with deployment, run ID, and `issue_review` boundary; the prepared-workflow action is available in the footer. |
| PASS | 1280x720 | replay | Scene 11 -> Scene 12 | Click `Submit`. | The route changes to `resume`; the output shows `submitted`, `approved: true`, selected issue `risk-1`, and a created issue. |
| FACTUAL | 1280x720 | replay | Scene 11 -> Scene 12 | Reload approval and click `Request revision`. | The route changes to `resume` and shows `cancelled`, `approved: false`, no selected issues, and a revision report. The replay uses `run_recorded_lda_report_revision`, so it does not preserve the submitted branch's run ID despite the same-run wording. |
| PASS | 1280x720 | replay | `/present#scene/prepared-lifecycle/{discover,draft,diagnose,repair,artifact,deployment}` | Open each Scene 8 beat directly. | The phase rail advances through Discover, Draft, Diagnose, Repair, Artifact, and Deployment; staged chat groups and factual evidence change with each beat. |
| PASS | 1280x720 | replay | `/present#scene/run-from-deployment/operation` | Open the Scene 9 operation beat. | `workflow.runs.start` is shown as interrupted with deployment, run ID, and `issue_review` boundary; the prepared-workflow action is available in the footer. |
| PASS | 1280x720 | replay | Scene 10 -> Scene 11 | Click `Submit`. | The route changes to `resume`; the output shows `submitted`, `approved: true`, selected issue `risk-1`, and a created issue. |
| FACTUAL | 1280x720 | replay | Scene 10 -> Scene 11 | Reload approval and click `Request revision`. | The route changes to `resume` and shows `cancelled`, `approved: false`, no selected issues, and a revision report. The replay uses `run_recorded_lda_report_revision`, so it does not preserve the submitted branch's run ID despite the same-run wording. |
| PASS | 1280x720, 1024x768 | replay | `/present#scene/resume-output-evidence/trace` | Open the trace beat after the decision branches. | Recorded execution frames render, including `review_issues` interrupt/continuation and `end_cancelled`; the UI exposes the evidence inspector. |
| PASS | 1024x768 | fallback | `/present#scene/typed-human-boundary/approval` | Set the presentation target to `http://127.0.0.1:1/rpc` and reload. | The approval route remains usable with the replay-backed payload and decision form; no live-ready badge is shown. The browser session target was restored afterward. |
| PASS | command line | live health | `/api/health` | Request the web-server health endpoint. | `200` with `{"ok":true,"status":"ok"}`. |
| PASS | command line | live health | `/api/connect` -> `http://127.0.0.1:8765/rpc` | POST the configured RPC target to the web server. | `workflow.health` returned `status: ok`, a store root, and equivalent CLI `uv run wf status`. |
| BLOCKED | 1280x720 | live end-to-end | Scene 10 -> Scene 12 | Start the prepared run, submit the approval form, and inspect output/trace against the live server. | Live health is verified, but a complete stateful live browser path was not completed in this rehearsal. Do not claim live output or trace success. |
| BLOCKED | 1280x720 | live end-to-end | Scene 9 -> Scene 11 | Start the prepared run, submit the approval form, and inspect output/trace against the live server. | Live health is verified, but a complete stateful live browser path was not completed in this rehearsal. Do not claim live output or trace success. |
### Operator Interpretation
@@ -45,7 +45,7 @@ the defense unless the full stateful live path is rehearsed separately.
1. Decide whether the revision branch should preserve the submitted branch's
run identity or be presented as a separate prepared recording.
2. With the example server running, start from the Scene 10 footer action and
2. With the example server running, start from the Scene 9 footer action and
record the live run ID, approval payload, submitted/revision-requested
outcome, output, and trace frames.
3. Keep the unavailable-target fallback check in the pre-defense checklist so
@@ -32,12 +32,12 @@ can be shown after selecting replay with the runbook's session-storage switch.
| `architecture/api` | JSON-RPC handles transport concerns and delegates to WorkflowApi rather than owning domain behavior. | Architecture API node | Client operations | hidden | replay-only; Thesis System Architecture; `docs/project_map.md`; `docs/source_architecture.md` | `architecture/api` |
| `architecture/runtime` | Server composition supplies stores, provider projections, and the runtime while the core remains independent of MCP and Python behavior. | Runtime/providers focus | Architecture overview | hidden | replay-only; Thesis System Architecture; `docs/project_map.md`; `docs/source_architecture.md` | `architecture/runtime` |
| `agent-handoff/request` | I will now show a prepared demonstration built on this platform. The chat is a presentation interface, not the autonomous planner evaluated by the thesis. The chat translates a report request into the same public lifecycle operations an external agent could call. This prepared path demonstrates product behavior and recorded evidence, not a fresh model-performance result. | Agent request surface | Workflow substrate framing | hidden | replay-only; Constrained demo agent and prepared replay recipe | `agent-handoff/request` |
| `prepared-lifecycle/discover` | The later issue-review example first inspects configured local.lda_docs, report, and issue-board capabilities. | Prepared lifecycle: discovery | Capability/schema evidence | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/discover` |
| `prepared-lifecycle/draft` | It creates and edits a Draft for report generation, making the proposal visible before execution. | Prepared lifecycle: draft | Workflow definition | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/draft` |
| `prepared-lifecycle/diagnose` | Validation returns structured diagnostics, affected paths, repair hints, and suggested next actions. | Diagnostic receipt | Draft authoring surface | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/diagnose` |
| `prepared-lifecycle/repair` | These surfaces make invalid intermediate drafts repairable; they support a loop, but no hint guarantees success. | Repair guidance | Diagnostic receipt | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/repair` |
| `prepared-lifecycle/discover` | It inspects sources, capabilities, and schemas rather than guessing at hidden interfaces. | Prepared lifecycle: discovery | Capability/schema evidence | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/discover` |
| `prepared-lifecycle/draft` | Focused operations modify mutable authoring state before execution. | Prepared lifecycle: draft | Workflow definition | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/draft` |
| `prepared-lifecycle/diagnose` | Structured diagnostics identify the missing output projection. | Diagnostic receipt | Draft authoring surface | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/diagnose` |
| `prepared-lifecycle/repair` | One focused output-map edit resolves it; hints do not guarantee automatic repair. | Repair guidance | Diagnostic receipt | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/repair` |
| `prepared-lifecycle/artifact` | It saves the validated plan as immutable artifact lda_report_case_study version 1. | Prepared lifecycle: artifact | Lifecycle state | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/artifact` |
| `prepared-lifecycle/deployment` | This later issue-review example is richer than the thesis three-node deterministic report case study: it is an implementation extension built on the same platform. Deployment binds and validates a ready configuration but does not run the workflow. | Prepared lifecycle: deployment | Binding/readiness evidence | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/deployment` |
| `prepared-lifecycle/deployment` | Deployment binds and validates three local sources; execution starts in the next scene. | Prepared lifecycle: deployment | Binding/readiness evidence | hidden | replay-only; `examples/lda_report_workflow`; deployment inspect replay evidence | `prepared-lifecycle/deployment` |
| `run-from-deployment/input` | The deployment receives selected local documents and an issue-board path. | Workflow input panel | Prepared deployment | hidden | live-capable; `workflow.runs.start` replay evidence | `prepared-lifecycle/deployment` |
| `run-from-deployment/operation` | The public workflow.runs.start operation validates the deployment and input, creates a persisted Run, and begins the reusable graph. | Start operation | Workflow graph | hidden | explicit run; `workflow.runs.start` replay evidence | `run-from-deployment/operation` |
| `run-from-deployment/graph` | The graph reads documents, analyzes them, builds a report, drafts proposed issues, and pauses at a declared review interrupt before issue-board changes. | Workflow graph | Run operation receipt | hidden | live-capable; `workflow.runs.start` replay evidence | `run-from-deployment/operation` |
+27 -33
View File
@@ -5,7 +5,7 @@ Audit date: 2026-07-13. This review uses the complete
[rehearsal log](presentation-rehearsal-log.md), the
[presentation runbook](defense-presentation.md), the
[Q&A runbook](defense-qna.md), and the locally available private narrative
notes in `random shit/`. The matrix covers all 14 scenes; the log records
notes in `random shit/`. The matrix covers all 13 scenes; the log records
targeted replay/fallback checks and does not establish a completed live
end-to-end run.
@@ -42,54 +42,48 @@ end-to-end run.
- Next action: keep
### Scene 6 — Architecture Zoom
- Audience takeaway: The same public surface reaches API, runtime/providers, stores, and a typed NodeUse execution.
- Visible proof: Client, API, runtime, and NodeUse focus beats provide a semantic zoom with package and operation evidence.
- Missing or confusing: Four nested levels risk becoming a component tour unless each level answers what responsibility moves inward.
- Audience takeaway: The same public surface reaches API, runtime/providers, stores, and the execution kernel.
- Visible proof: Client, API, and runtime focus beats provide a semantic zoom with package and operation evidence.
- Missing or confusing: Three nested levels risk becoming a component tour unless each level answers what responsibility moves inward. NodeUse remains an optional deep dive.
- Next action: keep
### Scene 7 — Author, Validate, Repair
- Audience takeaway: Discovery and structured diagnostics reduce agent guessing during workflow authoring.
- Visible proof: Discover, author, diagnose, and repair beats show the operation loop and repair guidance.
- Missing or confusing: The repair result should be verbally connected to the valid artifact, not treated as another isolated tool call.
- Next action: keep
### Scene 8 — Agent Request
### Scene 7 — Agent Request
- Audience takeaway: A thin external-agent surface can request work without pretending to execute it.
- Visible proof: Send reveals the prepared conversation and four discovery tool calls with no run claim.
- Visible proof: Send reveals the prepared conversation and discovery tool calls with no run claim.
- Missing or confusing: The chat surface could be mistaken for the thesis product unless the substrate remains the stated center.
- Next action: keep
### Scene 9 — Prepared Workflow Lifecycle
- Audience takeaway: The request is translated into Discover, Draft, Validate, Artifact, and Deployment stages.
- Visible proof: The phase rail, staged messages, and changing factual evidence advance across all five beats.
- Missing or confusing: Deployment records a local run request but does not execute; that boundary needs an explicit spoken pause.
### Scene 8 — Prepared Workflow Lifecycle
- Audience takeaway: The request becomes six concise authoring stages: Discover, Draft, Diagnose, Repair, Artifact, and Deployment.
- Visible proof: The phase rail, staged messages, structured diagnostic, focused output-map edit, and changing factual evidence advance across all six beats.
- Missing or confusing: Deployment binds and validates three local sources but does not execute; that boundary needs an explicit spoken pause.
- Next action: keep
### Scene 10 — Run From Deployment
### Scene 9 — Run From Deployment
- Audience takeaway: Execution starts from a ready deployment through a public run operation.
- Visible proof: Input, operation, and graph beats show selected inputs, `workflow.runs.start`, a run ID, and the typed boundary.
- Missing or confusing: The live end-to-end path was blocked, so live output must not be implied from the replay-backed operation view.
- Next action: factual fix
- Next action: keep
### Scene 11 — Typed Human Boundary
### Scene 10 — Typed Human Boundary
- Audience takeaway: A persisted run pauses at a typed issue-review decision with explicit submitted and revision-requested outcomes.
- Visible proof: Interrupt payload, selected issue, comment field, and both decision controls are visible in replay.
- Missing or confusing: The revision replay uses `run_recorded_lda_report_revision`, so “same run” wording is false for that branch.
- Next action: factual fix
### Scene 12 — Resume, Output, Evidence
### Scene 11 — Resume, Output, Evidence
- Audience takeaway: A decision leads to inspectable output and trace evidence for the workflow run.
- Visible proof: Resume, output, and trace beats show status, report/issue result, interrupt continuation, and terminal frames.
- Missing or confusing: Submitted replay continuity is demonstrated; live continuity and revision same-run continuity are not yet established.
- Next action: factual fix
### Scene 13 — Evaluation
### Scene 12 — Evaluation
- Audience takeaway: The 36 trials are bounded engineering evidence about operability and UX failure modes.
- Visible proof: Cohort, validity, and findings beats separate audited evidence from benchmark-style claims.
- Missing or confusing: The evaluation arrives after demo evidence, so the presenter must state that failures motivate the product-surface argument.
- Next action: keep
### Scene 14 — Limits and Conclusion
### Scene 13 — Limits and Conclusion
- Audience takeaway: The contribution is a useful substrate, not a production agent, scheduler, or broad benchmark.
- Visible proof: Limits, future, conclusion, and questions beats distinguish implemented core from future layers.
- Missing or confusing: Q&A must not begin before the contribution sentence and limitations have landed.
@@ -101,19 +95,19 @@ end-to-end run.
- Scene 1 -> 2 works: the goal becomes the missing reusable-automation contracts.
- Scene 5 -> 6 works if Deployment is the handoff: lifecycle vocabulary becomes the architecture that owns it.
- Scene 7 -> 8 works: authoring/repair operations become a thin external-agent request surface.
- Scene 8 -> 9 works and is not a duplicate: request/discovery becomes staged lifecycle evidence.
- Scene 9 -> 10 is the critical boundary: Deployment prepares; Scene 10 alone starts execution. Keep the explicit no-run wording.
- Scene 10 -> 11 -> 12 works as one climax in replay, but the revision branch has a factual run-identity defect and the live path is blocked.
- Scene 13 -> 14 works: evaluation limits become the contribution and future-work close.
- Scene 6 -> 7 works: architecture hands off to a bounded prepared request surface.
- Scene 7 -> 8 works and is not a duplicate: request/discovery becomes staged lifecycle evidence.
- Scene 8 -> 9 is the critical boundary: Deployment prepares; Scene 9 alone starts execution. Keep the explicit no-run wording.
- Scene 9 -> 10 -> 11 works as one climax in replay, but the revision branch has a factual run-identity defect and the live path is blocked.
- Scene 12 -> 13 works: evaluation limits become the contribution and future-work close.
### Order, Duplicated Beats, And Q&A
The 14-scene order is coherent: motivation, positioning, boundary, vocabulary,
architecture, authoring, then demo proof, evaluation, and limits. No duplicated
demo beat was found. Scene 8 introduces the request, Scene 9 owns authoring and
deployment, Scene 10 owns run activation, Scene 11 owns the decision, and Scene
12 owns resume/output/trace. The Q&A branch belongs after Scene 14's Questions
The 13-scene order is coherent: motivation, positioning, boundary, vocabulary,
architecture, request, authoring, then demo proof, evaluation, and limits. No duplicated
demo beat was found. Scene 7 introduces the request, Scene 8 owns authoring and
deployment, Scene 9 owns run activation, Scene 10 owns the decision, and Scene
11 owns resume/output/trace. The Q&A branch belongs after Scene 13's Questions
beat; opening a prepared branch earlier would interrupt the argument and make a
discussion answer look like core evidence.
@@ -122,7 +116,7 @@ discussion answer look like core evidence.
- **Factual:** The revision-requested replay has a separate run ID while the
surrounding same-run wording suggests continuity. Correct the recording or
label that branch explicitly as a separate prepared recording.
- **Factual:** The live health boundary passed, but live Scene 10 -> 12 was
- **Factual:** The live health boundary passed, but live Scene 9 -> 11 was
blocked. Do not present live output or trace as rehearsed evidence.
- **Visual:** No visual defect was recorded in the available rehearsal log. The
matrix remains the acceptance checklist for both `1280x720` and `1024x768`.
+9 -8
View File
@@ -69,27 +69,28 @@ pnpm --filter @lda/console build
Screenshots require human approval and are not pixel-diff tests.
## Scene 9 Staged Message Review
## Scene 8 Staged Message Review
Review these five deep links at `1280x720` and `1024x768`, at 100% zoom:
Review these six deep links at `1280x720` and `1024x768`, at 100% zoom:
- `/present#scene/prepared-lifecycle/discover`
- `/present#scene/prepared-lifecycle/draft`
- `/present#scene/prepared-lifecycle/validate`
- `/present#scene/prepared-lifecycle/diagnose`
- `/present#scene/prepared-lifecycle/repair`
- `/present#scene/prepared-lifecycle/artifact`
- `/present#scene/prepared-lifecycle/deployment`
Confirm that one labelled textarea remains visible in every phase, is empty in
Discover and Validate, and contains the exact prepared prompt in Draft,
Discover, and contains the exact prepared prompt in Draft, Diagnose, Repair,
Artifact, and Deployment. Confirm the right phase projection remains dominant,
the panes do not overlap, and `document.documentElement.scrollHeight` equals
`clientHeight`. Reload direct hashes before capture: in dev mode, immediate
screenshots during an in-place hash transition can catch the presentation
animation between surfaces even though the settled/reloaded route is correct.
Edit and Send in Draft, then verify Validate contains that exact edited user
turn. Repeat from Artifact to Deployment. On Deployment, Send must show only
Edit and Send in Draft, then verify Diagnose and Repair contain that exact edited
user turn. Repeat from Artifact to Deployment. On Deployment, Send must show only
`Run request prepared for the next execution slice.`; compare `/api/rpc`
request counts before and after to confirm no execution call was made. Scenes
1012 own the subsequent run activation and live execution path; do not use
Scene 9 to claim a run has started.
911 own the subsequent run activation and live execution path; do not use
Scene 8 to claim a run has started.