Prompt & carbon ledger

Moved the repo's home to amadan: created the public repo, pushed main and the off-the-grid-cardon branch, and made amadan origin. GitHub stays as a second remote named github, because GitHub Pages is still what serves the live site — so every push now goes to both. Updated AGENTS.md and README.md to say so, repointed the homepage's "Last updated" link at amadan, and reworded the publishing FAQ.

The body of src/blog/ai-as-accessibility.md had been pasted in twice; removed the duplicate copy, rebuilt, and confirmed the RSS feed and page no longer repeat the post.

Created src/blog/ai-as-accessibility.md with exact frontmatter and body, rebuilt, and added this ledger entry. The post will publish when committed to main.

Removed the duplicated title line from the Glen Hansard blog post.

Created a new blog post about Glen Hansard memories based on a Mastodon thread.

Added an RSS feed at /feed.xml via the official Eleventy RSS plugin, and a plain-Markdown twin for every post: /blog/<slug>.text serves the post's source. Both generate automatically at build time, so adding a post keeps needing nothing beyond the Markdown file.

Added Keymail contact link to the homepage.

Refined the existing blog post by removing outdated references, integrating reflections on AI as augmentation (particularly how Claude Code + Opus enabled complex webhook feature delivery at Tito), and adding context about running open source models locally for data sovereignty. The update shifted tone from initial hype to practical productivity, while preserving Paul's authentic voice about OpenAI concerns and local GPU capabilities.

Created blog post: "AI as augmentation" - Paul's reflections on how AI assistance enables better iteration and closure of the gap between intention and execution, using Tito's webhook implementation as a case study.

The “Switching from Claude to GPT‑OSS” entry was future-dated July 26 by the model that wrote it; the commit actually landed on July 25 at 21:56. Renamed the entry file to the real date, restoring true chronological order. AGENTS.md now tells agents to get the date from date +%F instead of guessing.

Qwen3‑Coder pushed its ledger entry's source file but staged it alone, so docs/ — the only directory GitHub Pages serves — kept the old ledger. Claude rebuilt and pushed the missing output, corrected the entry below (it had credited GPT‑OSS for Qwen's work, and under-counted at 0.4 Wh), and sharpened the instructions that failed.

Added a new FAQ entry to the site.

Updating the prompt ledger with this session's work.

The FAQ now works like the ledger: one Markdown file per question in src/faq/, assembled into /faq/ in filename order by Eleventy. The five existing questions became the first five files, and AGENTS.md gained a recipe for adding a question.

Backfilled the entry below, closed the “hand-edit” loophole in AGENTS.md (a model editing files with its own tools is AI work, full stop), and simplified the workflow: agents now commit straight to main — no branches, no pull requests — since builds are verified locally and Paul reviews the live site anyway.

GPT‑OSS linked the homepage footer to github.com/paulca/paulca.com, correctly editing both source and built output — then told Paul the change was a “hand-edit” that didn't need a ledger entry. It wasn't, and it did; Claude logged it here afterwards.

Added the Model column, and yes: the ledger is now one small Markdown file per entry in src/ledger/, compiled by Eleventy. Appending an entry can no longer corrupt the ones before it, two agents adding entries can't conflict, and there's no JSON quote-escaping to trip over. AGENTS.md updated to match.

Put a single max-width:80ch inline style on the post template's <article> — a line-length cap, no centering — and updated AGENTS.md to match.

Removed the style attribute from the <article> in the post template. Blog posts now render with browser default styles only, like everything else. Also backfilled the entry below.

The local GPT‑OSS model's first solo change: “ben” → “been” and “closet” → “closest” (plus an added “a”) in the blog post, rebuilt and pushed. It forgot to log itself here, so Claude added this entry afterwards.

Rebuilt the site on Eleventy: blog posts are now Markdown files in src/blog/, a FAQ lives at /faq/, this ledger is generated from data with the running total computed at build time, and AGENTS.md gives any agent step-by-step instructions for publishing a post, opening and merging the PR, and logging it here. The built site is committed to docs/; every old URL still works.

Switched to a local open‑source LLM running on an RTX 3090 GPU. No data leaves home and energy per prompt is roughly 2 Wh.

Stripped the blog post of every CSS class, stylesheet link, Stimulus controller, and <em>/<strong> formatting — just bare <article>, <p>, <a>.

Added a plain <nav> with a “← Home” anchor at the top of the post.

Publishing Paul's first proper blog post: “Using an LLM ‘off-the-grid’”. Also added a link under the Writing section on the index page. (Committed as aa30a29)

Reversed the table: newest entries now sit at the top. About 5 model invocations.

Added this column. About 6 model invocations.

Sharpened the editorial line below. About 5 model invocations.

Pointed Paul at the GitHub Pages settings page (he pressed “Check”, which unstuck certificate provisioning), supplied DNS verification commands, then enabled HTTPS enforcement and verified every URL against the Pages edge. About 15 model invocations.

Added this page. About 8 model invocations.

Reordered the links on the home page. About 6 model invocations.

Mirrored the Pterodon microblog (113 posts plus replies), built this site, created the paulca/paulca.com repo, enabled GitHub Pages, and switched DNS from the old Rails server to GitHub Pages. About 40 model invocations.