Skip to content

Documentation update log

2026-09-11

  • Update: Agent discoverability, after a Claude.ai tutoring session could not reach the Markdown twins (its fetch tool only opens addresses seen as links in prior results, and the twins were code spans and head-only tags). Every rendered page now has a "View this page as Markdown" button beside "View source" and a "Machine-readable versions" line at the end of the article linking the Markdown twin, the raw GitHub source, llms.txt, and llms-full.txt; the site footer links llms.txt, llms-full.txt, and the agent guide; llms.txt lists the HTML, Markdown-twin, and raw-source address of every page, states the site-path to docs/*.md mapping, and gives the corpus size; For AI agents uses absolute links and gained "If you cannot fetch this site"; Learn with an AI tutor links every address and adds a raw-source column; the landing page links the indexes in body text. Verified live: the Markdown twins, section-listing twins, llms.txt, llms-full.txt, sitemap.xml, and robots.txt all return 200.
  • Update: Lesson 2 and Lesson 3 follow the Lesson 1 pattern: each is now a text-only 50-minute in-person lecture (Lesson 2: the data life cycle as a definition list, FAIR and CARE, a table of the 2026 NIH, NSF, and SRP plan requirements, repositories and licenses, and a draft-the-NIH-2026-plan exercise; Lesson 3: bias sources and LLM failure modes, the NIH, NSF, and journal rules, the never-paste list, agent practices, and two of the six scenarios), and the full material moved to new self-paced pages, Lesson 2 homework and Lesson 3 homework, each with twelve modules and checkpoints, lesson: metadata, "In brief" summaries, key-terms glossaries, and text descriptions for the data life cycle, RO-Crate, FAIR, and Dartmouth figures. Content tabs in Lesson 2 (data sources, licenses) became definition lists so linear readers and AI parsers see every option. Two quiz questions were added (the NSF versus NIH plan formats; evidence for an agent's claims).
  • Update: Lesson 1 is now a 50-minute in-person lecture (definition, six pillars as a definition list, a Gold Standard Science tenet-to-practice table, three compliance facts, a six-question activity, three quiz questions, key terms), and its full material moved to a new self-paced homework page, Lesson 1 homework: Open Science, self-paced, organized as twelve modules with checkpoint questions. New Gold Standard Science content in both: the nine tenets of Executive Order 14303, the June 2025 OSTP guidance and its 1 September annual reporting cycle, NIH's 22 August 2025 implementation plan and the HHS 2025 and 2026 reports, the SPARC brief on how agencies tie the tenets to public access, OSTP's July 2026 "Science: A New Golden Age", and the debate over political oversight. Publication-cost facts refreshed as of September 2026 (NIH cap still not final; OMB 2 CFR 200.461 still a proposal).
  • Creation: Learn with an AI tutor: how to load a lesson into Claude, ChatGPT, Gemini, or NotebookLM, with prompts for lecture, tutor, and interactive modes and accommodation prompts for screen-reader users, deaf learners, and learners whose first language is not English. For AI agents gained a "Teaching a lesson" section.

  • Creation: Accessibility statement (WCAG 2.2 AA target, features, use with assistive technology and AI assistants, known limitations, reporting). Lesson 1 pages carry a machine-readable lesson: frontmatter block (objectives, key terms, duration, delivery modes, accessibility profile); the build now injects a schema.org LearningResource JSON-LD record and okf:lesson-* meta tags into every Lesson page. Lesson 1 figures have text descriptions, both pages have "In brief" plain-language summaries and glossaries, and the stylesheet adds visible focus outlines, reduced-motion support, and a screen-reader-silent external-link marker.

  • Update: Moved the repository to the UNM CARC organization (UNM-CARC/dust-2026) and the site to https://unm-carc.github.io/dust-2026/; every site and repository link was updated. Links to the 2025 site are unchanged.

  • Update: Added the Texas A&M Superfund Research Center as a third supported center alongside the University of Arizona DUST Center and the UNM METALS Center on the landing page, About this training, Credits, and Additional resources, and in the site description and repository README. Lesson examples still pair Arizona and New Mexico items; Texas examples are a follow-up.
  • Creation: Built DUST 2026 from DUST 2025 (commit 29027db, 2025-10-29), moving from MkDocs Material to Zensical, structured as an Open Knowledge Format (OKF v0.2) bundle, and restyled after the UNM CARC documentation. Every page carries provenance frontmatter (generated, sources with per-file last_modified) and a source footer; all rewritten pages are unverified pending the author's review. Contact and repository details moved to UNM (tswetnam@unm.edu, UNM-CARC/dust-2026).
  • Creation: Reorganized the site into Lessons 1–3 and About (training overview, resources, credits, guidance for AI agents, this log) with kebab-case URLs; MIGRATION.md in the repository maps every 2025 URL to its 2026 location. Wrote the landing page, section listings, and For AI agents, and added the llms.txt / llms-full.txt agent indexes, the Markdown mirror, okf:* meta tags, and robots.txt.
  • Update: Reframed the whole training for a joint audience: the University of Arizona DUST Center and the UNM METALS Center. Every "DUST Example" became an "SRP Example" pairing an Arizona item (arsenic, mine tailings, phytoremediation, lung injury) with a New Mexico item (uranium and metal mixtures from abandoned mines, Navajo Nation and Pueblo of Laguna partners). Examples name centers, partner communities, and project topics, never individual investigators.
  • Update: Lesson 1 now carries the September 2026 policy landscape: zero-embargo public access in force at NIH (1 July 2025), DOE, EPA, USGS, NSF, and USDA while the 2022 OSTP memo is under repeal; the pending NIH APC cap and the proposed OMB 2 CFR 200.461 change; Gold Standard Science (EO 14303, the Kratsios guidance, agency plans); the FY2026 NIEHS and SRP appropriation; 2026 article-processing charges; Diamond OA and the cOAlition S pivot; openRxiv and arXiv governance; two new quiz questions.
  • Update: Lesson 2 now teaches the NIH 2026 data management and sharing plan format (NOT-OD-26-046, NOT-OD-26-100), the NSF PAPPG 24-1 supplements and Data Management and Sharing Plan, and the SRP Data Management and Analysis Core requirement; adds "When the data disappear: EJScreen and data rescue (2025)" and a Dependency self-assessment; expands CARE with a "Research on Navajo Nation" warning (NNHRRB approval, chapter resolutions, data ownership, dissemination approval, UNM IRB guidance) and Local Contexts; refreshes repositories (ReDATA, UNM Digital Repository, Dryad, EDI, NIEHS CEBS, Tox Data Commons, EPA Science Inventory), metadata standards, cloud-native formats, and RO-Crate; replaces the DMP group exercise with a two-site Arizona and Navajo Nation metal-mixture scenario.
  • Update: Lesson 3 was rewritten for the agentic era: reasoning models and agents, NIH NOT-OD-25-132 on AI-written applications, NIH's May 2026 fabrication and plagiarism framing, the NIH and NSF bans on AI in peer review, journal policies (ICMJE, COPE, Springer Nature, Elsevier, PLOS), the July 2025 hidden-prompt scandal, LLM-specific failure modes, a new "Agentic AI and Research Integrity" section, current energy and water figures with a Southwest data-center angle, the September 2026 regulatory landscape (EU AI Act and Digital Omnibus, US executive orders, state laws), consumer-versus-enterprise account guidance, Superfund-specific privacy rules covering tribal data, six discussion scenarios (two new: a Navajo Nation household survey and the hidden prompt), an agent checklist, and a new quiz.
  • Update: About this training, Credits and attribution, and Additional resources rewritten for 2026: joint-audience framing, Version 2.0 history, credits for DUST 2025, the UNM CARC FOSS edition, GPT 101, Zensical, OKF, and the CARC design; resources gained sections on federal data preservation, institutional repositories, Indigenous data governance, cloud-native formats, scholarly indexes, environmental-health metadata, and 2026 AI frameworks, and replaced the Academic Twitter entry with Bluesky and Mastodon.
  • Update: Repaired or replaced every dead link found in the 2025 site (AI Fairness 360, the Creative Commons chooser, GitHub Skills, FORRT for FOSTER, the CARE principles page, NIEHS worker training, EPA Science Inventory for ScienceHub, the Helsinki Ethics of AI course, the NSF and NIH policy hubs, GO FAIR, FAccT, and the old tswetnam GitHub username); corrected the misspelled OSTP director (Kratsios) and the Gold Standard Science fact-sheet link. Deliberate deviations: AI Fairness 360 links to its GitHub repository because the IBM host's TLS certificate has expired; UNESCO links are kept although unesco.org refuses automated connections.
  • Deletion: Dropped mkdocs.yml, the custom JavaScript (reading-time and print widgets; its keyboard shortcuts broke under a sub-path), the MathJax and polyfill.io scripts (no lesson uses math; polyfill.io was compromised in 2024), the University of Arizona-only theme, and seven unreferenced images (about 9 MB); downscaled the 12 MB FAIR-principles figure.

Machine-readable versions of this page: Markdown twin · raw source on GitHub · llms.txt · llms-full.txt (whole site). See For AI agents.