GCC2026 — Slide-by-slide Outline

Companion to CONTEXT.md. Each slide entry: tier (MVP / Polish / Stretch), goal, body sketch, MVP fallback.

Implementation lives at ~/projects/repositories/jmchiltondotnet/src/content/slides/gcc2026.mdx. Sections in MDX correspond 1:1 to entries below.

Pacing target

8 minutes presenting, 2 minutes Q&A. Roughly 16–22 slides total at 20–30s each plus two video blocks (~3 min each).

Block A — Thesis (≈2:00)

A1. Title — ~10s — Tier 0

Goal: establish credibility and tone. Body: Title (working: “Reproducibility in the Age of Agents”), subtitle (“Designing for human reproducibility, accelerating it for everyone else”), name, GCC2026. Notes: terminal-aesthetic title slide matching prior decks.

A2. The new crisis — ~25s — Tier 0

Goal: state the problem with no hedging. Body: “Agent-driven science is scaling faster than our ability to trust it.” Volume metric or imagined headline. The phrase “we’re about to automate the reproducibility crisis” lands here. MVP fallback: text-only slide.

A3. The temptation to adapt away — ~25s — Tier 0

Goal: name the wrong move so the right move lands. Body: brief acknowledgment that the easy move is to bolt chat onto Galaxy or build agent-only shortcuts that bypass provenance. Frame as the slop path. MVP fallback: text-only slide; could be a one-line aside if pacing tight.

A4. The claim — ~30s — Tier 0

Goal: the headline argument of the talk. Body: “The infrastructure built for human reproducibility is exactly what agents need.” Galaxy should double down, not adapt away. Designing for humans accelerates AND incentivizes reproducibility for agents. Notes: this is the slide whose sentence the audience should remember if they remember nothing else.

A5. Two case studies — ~20s — Tier 0

Goal: set up Block B and Block C. Body: “Two new ways to construct the same Galaxy workflow against the same scientific question — one starting from a documented history, one starting from a conversation. Both leaning on the same reproducibility primitives.” Body, line 2: introduce the MRSA mobile-AMR comparative-isolate application (one-liner from galaxy-brain#12). MVP fallback: include a single icon for each case study + the MRSA one-liner.

Block B — Case Study 1: History → Workflow (≈3:00)

B1. Setup — History Notebooks — ~30s — Tier 0

Goal: introduce the substrate before showing the demo. Body: “History Notebooks: Galaxy-flavored markdown attached to histories. Datasets, parameter tables, visualizations embedded. Every revision attributed (user / agent / restore). Already in the codebase on history_pages.” 1 screenshot showing notebook with embedded dataset display. MVP fallback: static screenshot only.

B2. The pivot — extraction from narrative — ~15s — Tier 0

Goal: name the new mechanism the demo shows. Body: “Once a notebook references the outputs that matter, Galaxy walks the provenance graph backward and proposes a workflow.” Point at issue #22709. Notes: keep this terse; the video carries the explanation.

B3. Demo video 1 — ~2:30 — Tier 1 (Tier 0 fallback below)

Goal: the audience sees the mechanism end-to-end. Source: edited cut of MRSA history → notebook narration → graph-backed extraction → workflow with seeded report. Storyboard in DEMO1_STORYBOARD.md. Tier 0 fallback: sequence of 4–6 still screenshots advanced with reveal.js fragments; jmchilton narrates live. Tier 1 target: ~3 min YouTube embed, sourced from a real (or convincingly seeded) MRSA history. Tier 2 stretch: the video is from a real biological run, not seeded data.

B4. What just happened — ~15s — Tier 0

Goal: transfer takeaway from video to argument. Body: “Narrative drove extraction. The notebook became the workflow’s report. Same provenance graph the history already had — agents didn’t need a new substrate.” Pointer to galaxy-notebooks paper (vault link / preprint URL when available).

Block C — Case Study 2: Conversation → Workflow (≈3:00)

C1. Setup — the other direction — ~20s — Tier 0

Goal: contrast with B1. Body: “Same application. No Galaxy server in the loop yet. We’re going to build the workflow before we ever run it — and we’re going to validate it the whole way.” Mention Format2 as the writable artifact and gxwf as the static-validation core. MVP fallback: unchanged.

C2. Format2 + the 10,000-tool registry — ~25s — Tier 0

Goal: state the depth-of-validation differentiator. Body: one-paragraph version of the gxwf executive summary’s competitive claim: “Every workflow system validates something. Only Galaxy can validate the scientific tool invocation itself — because only Galaxy has 10,000+ typed tool schemas.” Mini side-by-side: a .ga tool_state blob vs. its Format2 equivalent. MVP fallback: text-only with the comparison table from WORKFLOW_STATE_EXECUTIVE_SUMMARY.md.

C3. Foundry — ~20s — Tier 0

Goal: introduce the agent-authoring layer above gxwf. Body: “Foundry compiles workflow-construction knowledge into agent skills. Each step (Mold) author → validate-via-gxwf → fix in a tight inner loop. Today: an interview-to-Galaxy pipeline.” One sentence is enough — the video shows it.

C4. Demo video 2 — ~2:30 — Tier 1 (Tier 0 fallback below)

Goal: show Foundry walking an MRSA-shaped intent into a validated Format2 workflow. Source: edited cut of agent interview → progressive Format2 build → gxwf diagnostics → corrected workflow. Storyboard in DEMO2_STORYBOARD.md. Tier 0 fallback: screencasted gxwf-validating-a-Format2-file-in-VS-Code only, with Foundry framed as forthcoming layer. Tier 1 target: ~3 min YouTube embed with the Foundry interview pipeline driving authoring. Tier 2 stretch: the Foundry-iterating-on-workflow visualization (see FOUNDRY_VIZ_IDEAS.md).

C5. What just happened — ~15s — Tier 0

Goal: parallel of B4. Body: “Validated against the same tool schemas the Galaxy UI uses. Same workflow as case study 1, reached from the other direction. Format2 is now a first-class authoring surface for humans and agents.” Pointer to gxwf paper and Foundry paper.

Block D — Synthesis & pointers (≈0:30)

D1. Two paths, one infrastructure — ~20s — Tier 0

Goal: the through-line. Body: simple diagram: MRSA story → (path A: history → notebook → graph extraction → workflow) + (path B: conversation → Foundry → gxwf → Format2 workflow) → one underlying provenance / schema substrate. Notes: if a clean visual doesn’t land, this slide can be a text triplet.

D2. The three papers + long video — ~10s — Tier 0

Goal: make the talk an ad for further reading. Body: three boxes with paper titles + the standalone long-video URL. Polish target: real DOI / preprint / repo URLs; QR code. MVP fallback: vault paths + “papers in preparation.”

D3. Thanks / Q&A — ~10s — Tier 0

Goal: prompt questions. Body: acknowledgments (Nekrutenko lab, Galaxy community, IWC, etc.), contact, GitHub.

Optional / Stretch slides (only if time and material both arrive)

S1. Foundry-iterating viz — Tier 2

Standalone slide / pre-roll for demo 2 showing the workflow growing step-by-step with gxwf gating each transition. See FOUNDRY_VIZ_IDEAS.md.

S2. Edit-source provenance — Tier 2

Single screenshot of a notebook revision panel with user/agent/restore badges. Embeds the “attribution as provenance” point if Q&A time allows.

S3. IWC corpus CI — Tier 2

One screenshot of an IWC PR with gxwf validate --strict running. Empirically grounds the depth-of-validation claim; only include if pacing has slack.

MDX skeleton mapping

The MDX file at src/content/slides/gcc2026.mdx should contain one <section> per slide above, in order, separated by ---. Use reveal.js fragments (<span class="fragment">…</span>) for incremental reveals inside A2/A4. Embed YouTube via <iframe> for B3 and C4; until those videos exist, embed a placeholder <img> for the storyboard contact sheet.