Posts
All the articles I've posted.
-
I fixed the LLM prompt rule twice, but the note should never have been split
An LLM-made sibling note kept coming out empty but for links. Two prompt fixes later, the source showed both courses shared one description: merge, don't split.
-
Reviewing OpenCode, I found only one of my three AI subscriptions works as-is
Before installing the open-source coding agent OpenCode, I checked my subscriptions: ChatGPT works as-is, Claude is barred by policy, Gemini access is gone.
-
Removing the embedding filter caught more duplicates, not fewer
I swapped an embedding filter plus LLM checks for one LLM review of all sibling notes to catch duplication. One hub-note rule took precision from 55% to 100%.
-
I asked what ACP is, and found two protocols with the same acronym
The ACP seen next to MCP is the Agent Client Protocol, linking editors and AI coding agents. IBM's same-acronym protocol merged into A2A; ACP v2 is in draft.
-
Switching embedding models flipped the ranking
Moving from OpenAI's text-embedding-3-small to large flipped the order of true duplicate paragraphs and look-alike sentences mixed at cosine 0.72-0.80.
-
LLM-generated sibling notes were copying the same paragraph word for word
Testing my Obsidian plugin on real articles exposed an LLM bug: sibling notes copied the same paragraph verbatim, and embedding duplicate detection missed it.
-
I verified on a different pair and never rechecked the original case
I fixed a false positive in my Obsidian plugin's LLM merge-judgment prompt and even closed the GitHub issues, but never retested the case that caused the false positive in the first place. Rerun days later under the real full-vault conditions, the judgment flipped, and actually performing the merge suggested the original call was likely wrong all along.
-
Of the 4 most list-heavy files, the subagent missed 2
A full audit I delegated to Claude Code subagents reported 2 of 56 files as CLEAN when they weren't. An A/B test on the prompt showed that on files dense with bullet lists, 'I read it all' meant a skim rather than an item-by-item check, and a prompt that forces a checklist fixed it.
-
Wrong, but 90% confident
I tried Laya, the open-source reproduction of TypeSafe's Jev, on my Obsidian plugin's duplicate-note decisions (Jev itself only through its docs, since my account isn't open yet). I measured CPU speed, 78% accuracy, and what fine-tuning would take, and the real problem turned out to be broken calibration: confidence around 0.9 on wrong answers.