CONTEXT

GCC2026 Talk — Context for Future Agents

Status: Active planning, MVP target ~1–2 weeks from 2026-05-23. North-star plan with explicit MVP/polish/stretch tiers so any of these slides can be presented even if downstream features slip.

Owner: jmchilton. Slide implementation lives in: ~/projects/repositories/jmchiltondotnet (Astro + reveal.js, MDX, route auto-publishes at /slides/gcc2026). New deck file: src/content/slides/gcc2026.mdx. Dev: make dev. Build: make build. Reveal.js version note: package.json declares reveal.js ^5.2.1 (CSS only, imported at SlideLayout.astro line 3); the actual JS runtime is loaded from CDN at reveal.js@4.5.0 (lines 181/185/190). Target v4.5.0 JS features until the layout is upgraded. All planning docs (this file, slide outline, demo storyboards, status logs) live here in vault/projects/history_markdown/gcc2026/.

Thesis (the two-minute version)

We’re about to automate the reproducibility crisis. Agents scale analysis faster than they scale trust. Galaxy’s response should not be to adapt away from its core principles; it should be to double down on them. Infrastructure designed for human reproducibility is exactly what agents need — and shipping it as agent-grade tooling accelerates and incentivizes reproducibility on both sides.

The case-study payoff: two new ways to construct Galaxy workflows that both lean on the same reproducibility infrastructure, demonstrated against the same scientific application.

Shared application — MRSA mobile-AMR comparative analysis

Both demos use the MRSA mobile-AMR demo vignette described in jmchilton/galaxy-brain#12:

Status (as of 2026-05-23): storyboard only. Real Galaxy history is expected to be ready in time but is not the gating risk — both demos can be presented from storyboard/seeded-history if the real run slips.

Presentation budget (10 min slot, 2 min Q&A reserved)

BlockTimeContent
Title + thesis~2:00”Automating the reproducibility crisis” + Galaxy’s role
Framing both case studies~0:15”Two new ways to construct the same workflow”
Case study 1 — notebook-driven extraction~3:00embedded heavily-edited demo of #22709 against MRSA history
Case study 2 — building the workflow without Galaxy~3:00embedded demo of Foundry interview→Galaxy pipeline + gxwf validation, same MRSA app
Synthesis + paper pointers~0:30three papers as citations; site as long-video pointer
Total presenting~8:00leaves ~2:00 for questions

Key referenced documents

Abstract & exec summaries (this directory)

Paper manuscripts (the three citations the talk advertises)

Implementation references

Slide infrastructure

Slide outline

The slide-by-slide outline (with MVP/polish/stretch markers and per-slide notes) lives in SLIDE_OUTLINE.md in this directory. Read that file alongside this one.

Demo videos

Two embedded YouTube videos, each ~3 minutes after heavy editing. Storyboards are in:

Note: demo 2 deliberately does not mention Nextflow conversion. The Foundry mode being demonstrated is the interview→galaxy pipeline (currently in development). Keep this consistent across all materials.

Tiered scope — MVP / Polish / Stretch

Every component is tracked at one of three readiness tiers so that if a feature slips, the talk can present a degraded-but-coherent version of the same material.

Tier 0 — MVP (present even if everything below slips)

Tier 1 — Polish (target for talk day)

Tier 2 — Stretch (nice-to-haves)

Parallel track — longer standalone thesis video

This is planned in parallel with the conference talk and shares assets but has its own timeline. Conference talk is essentially a “trailer” for this video and the three papers.

Goals:

Tracked in LONG_VIDEO_PLAN.md in this directory.

Risk register

RiskMitigation
MRSA history not run by talk dayUse seeded history from seed_demo_histories.py style; demos still functional.
#22709 / graph extraction merge slipsDemo 1 reverts to existing extraction surface; narrative still about “notebook-driven extraction” with graph view in storyboard.
Foundry interview→galaxy pipeline not demoableDemo 2 falls back to gxwf-only authoring in VS Code; Foundry framed as forthcoming layer. Still tells the “Format2 as agent-writable artifact” story.
Long video slips past talk dayTalk’s “long video coming” pointer becomes a “papers in preparation” pointer to vault drafts.
Slide infra changes neededExisting reveal.js setup is rich; no migration anticipated.

Open questions

Working agreements for future agents on this project