Lab Notes
What each AI harness changed on this site, how it did it, and what worked and what didn’t.
-
·
Lab note #47: Bake-off R2: Claude with Opus 5.5 High entry
The ask Round 2 of the bake-off: rewrite the AI & Automation page as a child of the Round 2 page, covering every harness I use, how they connect, how OpenClaw and Hermes remember, how I work, and every project already on the live page, using only the sources the task allows and nothing from…
-
·
Lab note #46: Round 2 setup: AI & Automation
The ask Set up Round 2: every harness writes the AI & Automation page about my harnesses and how I work, researching my Obsidian notes and interviewing my OpenClaw and Hermes agents, without publishing anything that poses a security risk. What changed How Claude (Cowork) over SSH: wrote the task and tools, created the round…
-
·
Lab note #45: Full-width text on bake-off round pages
The ask The Round 1 text had moved left but was still narrower than the results table. What changed How Claude (Cowork) over SSH: one rule added to the theme’s custom CSS in global styles, a class on the round page wrapper, re-rendered and checked with a headless browser screenshot. What worked, what didn’t The…
-
·
Lab note #44: One wide column for bake-off round pages
The ask Line the Round 1 text up with the wider results table, and make every round page look the same from now on. What changed How Claude (Cowork) over SSH: wrapped the generated round page in a wide group block and re-rendered Round 1. What worked, what didn’t Clean change, page returns HTTP 200.…
-
·
Lab note #43: Wider bake-off results and the DeepSeek entry
The ask Widen the cramped Round 1 results table and give OpenCode with DeepSeek v4.1 High credit for the entry it did publish. What changed How Claude (Cowork) over SSH: edited the bake-off engine and scoreboard script, re-ran the Round 1 report, checked the page. What worked, what didn’t The first re-run dropped every contestant…
-
·
Lab note #42: Models on Round 1 bake-off entries
The ask Show the model and effort setting behind every Round 1 contestant, keeping the same contestant names. What changed How Claude (Cowork) over SSH with WP-CLI and the bake-off engine. All five entry pages verified at HTTP 200. What worked, what didn’t Contestant IDs and page addresses stayed the same, so Git history and…
-
·
Lab note #41: Organized Harness Bake-off page
The ask Organize the Harness Bake-off page before Round 2 is added. What changed How Claude (Cowork) over SSH: edited the bake-off engine and protocol doc, re-rendered with WP-CLI, verified both pages return HTTP 200. What worked, what didn’t Clean run, no rollbacks. The Round 1 Claude and ChatGPT entries still have no model in…
-
·
Lab note #40: Automatic gallery counts and covers
The ask I added gallery photos and wanted the cards refreshed, with counts based on the actual galleries. What changed How Used SSH and WP-CLI, captured uploads, took snapshots, added a small site plugin and replaced the cards’ manual counts and covers with dynamic shortcodes. What worked / what didn’t Verified all six counts and…
-
·
Lab note #39: Update wheels and tactile photo cards
The ask I added photos to Wheels & Controls and Tactile Feedback and wanted the gallery cards updated. What changed Added linked photo covers and current counts to both cards: 25 photos for Wheels & Controls and 78 photos for Tactile Feedback. How Captured the uploaded photos in Git, took a snapshot and updated the…
-
·
Lab note #38: Space-efficient snapshots with retention
The ask I wanted backups to stop making a full new copy of every photo for each change and to keep storage growth under control. What changed Snapshots now share unchanged files while storing changed versions separately. Added automatic retention for recent, daily and weekly restore points, with a 2 GiB storage target and at…