update with the blog

Author: mdenolle

README.md

@@ -63,6 +63,8 @@ title: "Your Post Title"
 series: agents-for-academic-practice
 repo_url: "https://github.com/username/repo-for-this-post"
 skills_url: ""
+discussion:
+   url: "https://github.com/YOUR_REPO_OWNER/YOUR_REPO_NAME/discussions/123"
 ---
 Your post content here.
 ```
@@ -75,11 +77,12 @@ Your post content here.
 **Optional fields:**
 - `repo_url: "..."` – link to GitHub repository with related skills/instructions
 - `skills_url: "..."` – direct link to skills/instructions file
+- `discussion.url: "..."` – direct link to the GitHub Discussion for this post
 
 Posts with this series tag will automatically appear on `/blog/` with optional links rendered below each post.
 
 ### Add GitHub Discussions to your blog posts
-Readers can discuss each post directly on GitHub Discussions, with an optional giscus comment thread embedded in the post.
+Readers discuss each post on GitHub Discussions.
 
 #### Setup steps
 
@@ -88,104 +91,40 @@ Readers can discuss each post directly on GitHub Discussions, with an optional g
 - Scroll to "Discussions" and enable it
 - (Optional) Configure title and description for your discussion categories
 
-**2. Find your Discussion Category ID**
-To auto-create discussions for posts, you need the numeric category ID. Get it via GitHub GraphQL Explorer:
-
-1. Go to [GitHub GraphQL Explorer](https://docs.github.com/en/graphql/overview/explorer)
-2. Sign in with your GitHub account
-3. Paste this query, replacing `YOUR_REPO_OWNER` and `YOUR_REPO_NAME`:
-    ```graphql
-    query {
-       repository(owner: "YOUR_REPO_OWNER", name: "YOUR_REPO_NAME") {
-          discussionCategories(first: 10) {
-             nodes {
-                id
-                name
-             }
-          }
-       }
-    }
-    ```
-4. Click the play button. Copy the `id` of the category you want (e.g., "Blog Discussions" or similar).
-
-**3. Set your GitHub token**
-The automation script needs a GitHub Personal Access Token to create discussions on your behalf.
-
-1. Go to [GitHub Settings → Personal Access Tokens → Tokens (classic)](https://github.com/settings/tokens)
-2. Click "Generate new token (classic)"
-3. Give it a name like "Discussion Auto-Create"
-4. Check scopes: `repo`, `discussions`
-5. Generate and copy the token
-6. Save it as an environment variable: `export GH_TOKEN=ghp_xxxxxxxxxxxx`
-
-**4. Auto-create a discussion for a post**
-Once your post is created with the frontmatter above, run:
+**2. Create a discussion manually for the post**
+- In your repository, open the **Discussions** tab
+- Click **New discussion**
+- Use a title matching your blog post title (or close variation)
+- Publish the discussion and copy its URL
 
-```bash
-export GH_TOKEN=your_token_here
-python scripts/create_discussion.py _posts/YYYY-MM-DD-title.md
-```
-
-The script will:
-- Read your post frontmatter
-- Create a discussion on GitHub with the post title
-- Update your post file with the discussion URL in the frontmatter
-- Print success or error messages
+**3. Link the discussion in post front matter**
+Add this to the post front matter:
 
-**Example output:**
-```
-✓ Repository: marinedenolle/denolle.ai
-✓ Discussion created: https://github.com/marinedenolle/denolle.ai/discussions/42
-✓ Updated _posts/2026-02-26-welcome.md with discussion URL
+```yaml
+discussion:
+  url: "https://github.com/YOUR_REPO_OWNER/YOUR_REPO_NAME/discussions/123"
 ```
 
-#### Enable optional giscus comment embed (advanced)
-By default, posts link to GitHub Discussions but don't embed comments. To add a giscus-powered comment thread in each post:
-
-1. Find your repository ID via GraphQL:
-    ```graphql
-    query {
-       repository(owner: "YOUR_REPO_OWNER", name: "YOUR_REPO_NAME") {
-          id
-       }
-    }
-    ```
-    Copy the ID (e.g., `R_kgDOLN4fFg`).
-
-2. Find your Discussion Category ID (from step 2 above, e.g., `DIC_kwDOLN4fFg4CiPzp`).
-
-3. Edit `_config.yml` and uncomment/fill in the giscus section:
-    ```yaml
-    giscus:
-       enabled: true
-       repo: "YOUR_REPO_OWNER/YOUR_REPO_NAME"
-       repo_id: "R_kgDOLN4fFg"  # from GraphQL
-       category: "Blog Discussions"
-       category_id: "DIC_kwDOLN4fFg4CiPzp"  # from GraphQL
-       mapping: "pathname"
-       theme: "light"
-    ```
-
-4. Rebuild your site. New posts will now show a giscus comment thread below the discussion link.
+Once saved, the post will automatically show a **Discuss this post →** link and a discussion section.
 
-#### Troubleshooting
+#### Publish checklist (recommended)
+For each new blog post:
+1. Publish the markdown post in `_posts/`
+2. Create the matching GitHub Discussion
+3. Paste the discussion link into `discussion.url`
+4. Run `bundle exec jekyll build` to verify the site build
 
-**Scripts fails: "GH_TOKEN not found"**
-- Run: `export GH_TOKEN=your_token_here` before running the script
-
-**Script fails: "Discussions not enabled"**
-- Enable Discussions in your repository Settings (see step 1 above)
+#### Troubleshooting
 
-**Script fails: "Post file not found"**
-- Check the file path; it should be relative to the repository root (e.g., `_posts/2026-02-26-welcome.md`)
+**Discussion link does not appear in the post**
+- Check that front matter includes `discussion.url`
+- Check indentation under `discussion:` (two spaces before `url`)
 
-**Script fails: "Category not found"**
-- Verify category ID in _config.yml (or use GraphQL Explorer to get the right ID)
+**GitHub discussion not found / 404**
+- Verify the copied URL is correct and public to your readers
 
-**Discussion link appears but comment thread doesn't show**
-- Check that `giscus.enabled: true` in `_config.yml`
-- Verify `giscus.repo` and `giscus.repo_id` are correct
-- Rebuild the site
+**Discussions tab is missing in your repository**
+- Enable Discussions in your repository Settings (see step 1 above)
 
 ### Update styling
 Edit `assets/styles.css`.

_config.yml

@@ -41,13 +41,4 @@ media:
     # - title: "Panel: Climate-Hazard Intelligence at Scale"
     #   url: "https://denolle-lab.github.io/"
     # - title: "Lab update: Multi-sensor science and impact"
-    #   url: "https://denolle-lab.github.io/"
-# GitHub Discussions & Giscus configuration
-giscus:
-  enabled: false
-  repo: ""
-  repo_id: ""
-  category: "Blog Discussions"
-  category_id: ""
-  mapping: "pathname"
-  theme: "light"
+    #   url: "https://denolle-lab.github.io/"

_includes/discussion.html

@@ -3,44 +3,21 @@
   {% assign discussion = page.discussion %}
 
   {% if discussion.url and discussion.url != "" %}
-    <!-- Discussion link at top of post -->
     <div class="post-discussion-header">
       <a href="{{ discussion.url }}" target="_blank" rel="noopener noreferrer">
         Discuss this post →
       </a>
     </div>
 
-    <!-- Discussion section at bottom of post -->
     <section class="post-discussion">
       <hr>
       <h3>Discussion</h3>
 
       <p>
         <a href="{{ discussion.url }}" target="_blank" rel="noopener noreferrer">
-          Join the discussion on GitHub
+          Join the collegial discussion on GitHub
         </a>
       </p>
-      
-      {% if site.giscus.enabled and site.giscus.enabled != false %}
-        <!-- Embed giscus if enabled -->
-        <script 
-          src="https://giscus.app/client.js"
-          data-repo="{{ site.giscus.repo }}"
-          data-repo-id="{{ site.giscus.repo_id }}"
-          data-category="{{ site.giscus.category | default: 'Blog Discussions' }}"
-          data-category-id="{{ site.giscus.category_id }}"
-          data-mapping="{{ site.giscus.mapping | default: 'pathname' }}"
-          data-strict="0"
-          data-reactions-enabled="1"
-          data-emit-metadata="0"
-          data-input-position="top"
-          data-theme="{{ site.giscus.theme | default: 'light' }}"
-          data-lang="en"
-          data-loading="lazy"
-          crossorigin="anonymous"
-          async>
-        </script>
-      {% endif %}
     </section>
   {% endif %}
 {% endif %}

_layouts/default.html

@@ -30,7 +30,6 @@
           <div class="container prose post-shell">
             <p class="post-meta">{{ page.date | date: "%B %-d, %Y" }}</p>
             <h1>{{ page.title }}</h1>
-            {% include discussion.html %}
             {{ content }}
             {% include discussion.html %}
             <p><a href="{{ '/blog/' | relative_url }}">&larr; Back to blog</a></p>

_posts/2026-02-20-agentic-earth-update.md

@@ -2,7 +2,7 @@
 layout: default
 title: "AgenticEarth Update"
 series: agents-for-academic-practice
-repo_url: "https://github.com/mdenolle/marine-cv"
+repo_url: "https://github.com/mdenolle/https://github.com/mdenolle/academic-practice-agents"
 ---
 This update covers progress on agentic workflows for Earth systems analysis.
 

_posts/2026-02-28-introduction.md

@@ -0,0 +1,132 @@
+---
+layout: default
+title: "Introduction to the AI for Academic Practice Blog"
+series: agents-for-academic-practice
+repo_url: "https://github.com/mdenolle/academic-practice-agents"
+skills_url: ""
+discussion:
+  url: "https://github.com/mdenolle/denolle.ai/discussions/1"
+---
+
+## Practicing Science at the Speed of Foundation Models
+I am beginning a series of opinion essays on how generative AI is reshaping my scientific practice.
+
+Not as commentary from the sidelines.
+
+Not as technological evangelism.
+
+But as a working Earth scientist embedded in academia, research leadership, and family life.
+
+I am an earthquake scientist at the University of Washington. I build data-driven models of Earth systems. I am drawn to interdisciplinary problems where computational scale and physical reasoning intersect. I avoid incremental work whenever possible; I am interested in questions where better structure, better data, and better modeling genuinely change what we can know.
+
+I am also a mother of three young, energetic children. My spouse is a high-powered academic at UW as well. Our household runs on intellectual and athletic intensity, and calendar coordination. We are extraordinarily fortunate to have his parents deeply supporting this life. We are surrounded by remarkable Seattle families and friends who make this stage of life expansive and fun.
+
+I run. I ski. I mentor students seriously. 
+
+And over the last two years, I have intentionally integrated generative AI into nearly every dimension of my professional life.
+
+This blog is a record of that integration.
+
+## Why Write Now?
+
+The pace of foundation model development and the emergence of agentic ecosystems around them is too rapid to wait for the slow academic cycle.
+
+I cannot run controlled longitudinal studies on my own productivity. I cannot publish a meta-analysis on workflow acceleration. What I can do is experiment rigorously in my own practice:
+* Test systems.
+* Track what improves.
+* Discard what fails.
+* Refine prompts into structured agents.
+
+These essays will describe what actually works in scientific life and what does not.
+
+## My Operating Setup
+I use a business OpenAI account so that no model gets trained on my writing and others’ work I analyze through these tools. My work is organized through persistent
+
+“Projects” that gather multiple tasks for a broader purpose. Each major role in my academic life has its own structured AI environment:
+* Seismology course development
+* Communication statements
+* Large-scale proposal architecture (including a focused research organization centered on agents, foundation models, and Earth science)
+* Paper reviewing — my own manuscripts and those I referee
+* Curriculum design
+* Early-stage research concepts
+
+Each project has carefully written standing instructions and attached files as a knowledge base (e.g., a course syllabus, an NSF solicitation). Initially, I drafted the instructions informally. Then I iteratively refined them, asking the model to rewrite and formalize the instructions into a more agent-like specification, closer to structured task decomposition than casual prompting. The improvement in consistency and output quality was substantial. Instruction files have become an externalized role-task description that has lived in my brain for decades.
+
+For coding and lab development, I work in VS Code with Anthropic Sonnet or Opus, alongside premium GitHub Copilot. The annual cost is trivial relative to the leverage gained.
+Operationally, this has changed my workflow:
+* I no longer need additional TA time to polish documents.
+* My writing is clearer and structurally stronger.
+* Proposal framing is sharper and tailored to the RFP.
+* Departmental analytical work that once required a full day of searching and synthesis now takes under an hour (e.g., syllabus template compatible with university requirements). 
+* I prototype research software without needing to hire short-term RSE support.
+* I iterate privately before asking students or colleagues to invest time.
+
+All of my successful agent specifications are hosted publicly in [academic-practice-agents](https://github.com/mdenolle/academic-practice-agents). They are open for use, critique, and iteration. This is not proprietary infrastructure; it is evolving scientific workflow design.
+
+The overall cost: <$1000/year, cheaper than a single publication.
+
+## On Generalism and Expertise
+Foundation models are extraordinary generalists. But generalism is not expertise.
+
+In my own case, I am an A in geophysics, relative to the global distribution of practitioners. That depth comes from decades of training and research.
+
+Across many other domains of professional and personal life (proposal supporting docs, budgeting, visual design, communication framing, multi-sponsor pitching) I operate at a B, C, or sometimes D level.
+
+Base models reliably raise the floor. They do not turn me into an expert in those domains. But they elevate weak execution (D level) into competent execution (B level).  
+
+When enough peripheral tasks become competent and efficient, cognitive bandwidth returns: **I can think again**!
+
+That bandwidth can be reinvested into deeper work and into becoming A+ somewhere new.
+
+Used carefully, these systems do not replace domain expertise. They scaffold generalism. They allow experts to approximate polymathy by reducing the burden of incompetence in adjacent domains.
+
+For an academic, that is a transformational advantage.
+
+## What Has Changed
+The transformation is not that AI “writes for me.”
+The transformation is that I operate at a higher level more consistently.
+I:
+* Frame bigger research architectures.
+* Synthesize across fields more fluidly.
+* Move from idea to prototype faster.
+* Spend less time polishing language.
+* Spend more time mentoring students.
+* Build internal clarity before externalizing demands.
+
+I can do more.
+
+I can do it better and with intention.
+
+I rely less on short-term external support for routine tasks.
+
+And I free time to think more deeply.
+
+## Experts Wanted!
+Domain expertise has become even more important.
+Foundation models are powerful pattern synthesizers. They hallucinate and subtly distort scientific framing. Without deep subject-matter fluency, it is easy to be misled.
+
+In future posts, I will be explicit about:
+* Where models fail technically.
+* Where human judgment overrides outputs.
+* How context engineering improves reliability.
+* How it is impacting my multi-role productivity.
+
+## What to Expect
+Each post will examine one slice of academic life:
+* Teaching with AI assistance
+* Pre-reviewing manuscripts (starting with your own!)
+* Structuring large interdisciplinary proposals
+* Designing research agents
+* Departmental service workflows
+* Mentorship in the age of foundation models
+* Scientific identity under automation
+
+These posts are reflective field notes from active integration, not tutorials. They are open to discussion through GitHub Discussions, where you can add comments and suggestions for improvement.
+
+AI is already inside scientific life. The relevant question is not whether we adopt it. The question is whether domain experts shape its integration. This blog is my attempt to shape it, in public.
+
+I do not believe every scientist must integrate AI deeply into their workflow.
+
+But I do believe that ignoring its leverage entirely may quietly cost us time — time that could be reinvested into the parts of science we care about most.
+
+

blog.md

@@ -9,6 +9,7 @@ permalink: /blog/
     <p>A running collection of informal opinion pieces on how I use agents in my work life and professional development.</p>
     <p>These notes are spontaneous and practice-oriented, not peer-reviewed research outputs.</p>
     <p><strong>Posting format:</strong> each post is a single markdown page, and can include a link to the public GitHub repository containing skills/instructions files.</p>
+    <p><strong>Community note:</strong> to comment, open a post and click <em>Discuss this post →</em>. Discussions happen on GitHub to keep the conversation open, searchable, and collegial.</p>
 
     {% assign agent_posts = site.posts | where: "series", "agents-for-academic-practice" %}