Joshua Oliphant's Digital Garden

TIL: Shorter Embedding Body Improves Semantic Search Recall

Tue, 24 Mar 2026 00:00:00 -0000

[🌿 Budding] Last week I ran a set of autoloop experiments on my Obsidian vault recommender, comparing different text preparation strategies for embedding. I wanted to find out how much the input format affects recall when you're using sentence-transformers to surface related notes.

Taming AI-Generated Test Sprawl with Autoloop Coverage Analysis

Tue, 24 Mar 2026 00:00:00 -0000

[🌿 Budding] I ran pytest --co -q on ForkHub one morning and counted 489 tests. The project had maybe 12 source files. Something had gone sideways, I had started noticing this kind of test pileup in other projects too.

TIL: Wiki-links Are High-Quality Embedding Fuel

Mon, 23 Mar 2026 00:00:00 -0000

[🌲 Evergreen] Last week I was running autoloop optimization experiments on my vault recommender, trying to improve recall. I wrote about how a shorter embedding body improves semantic search recall. One of the variables I tested was what text gets sent to the embedding model. When I included wiki-links as a Related: field in the embedding input, recall@10 jumped from 0.91 to 0.97; the single largest improvement in fifteen experiments.

Hexagonal Agents: What If Your App's Business Logic Was an AI Agent?

Wed, 18 Mar 2026 00:00:00 -0000

[🌲 Evergreen] I was listening to a podcast (I think it was the Latent Space episode "Retrieval after RAG") and it got me thinking about JSONL files. I have all these JSONL files laying around from various projects that I'm not using, particularly from Claude Code usage. I've wanted to build event-driven agents for a while but kept struggling with what events to hook into. Somewhere in that thread of thought, I realized the architecture pattern I learned at my first engineering job maps onto agentic applications in a clean way.

Brooklet: The SQLite of Event Streaming

Sun, 15 Mar 2026 00:00:00 -0000

[🌿 Budding] I've been using Claude Code heavily since it came out, and one thing I noticed is that every session gets quietly recorded as a JSONL file:

Silencing peon-ping for Quick Claude Code Tasks

Sat, 07 Mar 2026 00:00:00 -0000

[🌿 Budding] I use peon-ping to get desktop notifications and sound alerts when Claude Code finishes a task. It's great for longer sessions where I context-switch while waiting — but it gets noisy fast when you're running /loop poll cycles that complete in seconds.

Vibe Code Alert - Andrej Karpathy on LLM Council

Sun, 30 Nov 2025 00:00:00 -0000

[🌲 Evergreen] Andrej Karpathy - LLM Council

Today I created my first Claude Agent skill

Sun, 19 Oct 2025 00:00:00 -0000

[🌿 Budding] Earlier this week Anthropic released Agent Skills. I've been reading about them all week, and today I finally built one myself. I've been using a tool called Mochi for spaced repetition, but not as much as I'd like to because it's a bit tedious to create cards, and I've been intending to create some AI tool or process to make it easier. So I created an agent skill for it, and even created my own plugin marketplace to store it. I'm sure I'll create more plugins.

Uv has a killer feature you should know about

Tue, 14 Jan 2025 00:00:00 -0000

[🌱 Seedling] Uv has a killer feature you should know about

What is in a DNS DKIM record?

Sun, 19 Apr 2026 01:29:39 -0000

[🌲 Evergreen] What is in a DNS DKIM record?

Quoting Stephen West

Sun, 19 Apr 2026 01:29:39 -0000

[🌲 Evergreen] What is happiness but a certain kind of harmony between a person and the life they lead?

Quoting Elizabeth Filips

Sun, 19 Apr 2026 01:29:39 -0000

[🌲 Evergreen] Instead of 1% improvements, aim for large leaps in progress (5%, 500%, or even 5000%) fueled by intense passion. This approach allows for faster learning and deeper focus, also known as deep work.

Prioritize deep work over incremental improvement

Sun, 19 Apr 2026 01:29:39 -0000

[🌿 Budding] This resonated with me because I also struggle to make incremental improvements. My natural tendency is to work hard on projects that I'm deeply interested in, and then likely not touch them for a long time (if ever). But the knowledge that I gain from working deeply like this tends to be very useful in other projects, conversations, etc.

Scaling Postgres Databases with Liquibase on Kubernetes

Wed, 12 Feb 2025 19:43:37 -0000

[🌲 Evergreen] I recently dived into scaling stateful applications in Kubernetes, focusing particularly on Postgres databases. My current project also integrates Liquibase for database migrations, which led me to explore its impact on scaling. Here are the consolidated notes from my research.

Markdown Link Content Scraper

Thu, 26 Dec 2024 00:00:00 -0000

[🌲 Evergreen] For a list of all my projects, visit my projects page.

Making programming more fun with an ai generated debugger

Tue, 24 Dec 2024 00:00:00 -0000

[🌲 Evergreen] Making programming more fun with an ai generated debugger

Making programming more fun with an ai generated debugger

Tue, 24 Dec 2024 00:00:00 -0000

[🌲 Evergreen] Geoffrey Litt

FOMO is not a strategy

Tue, 03 Dec 2024 00:00:00 -0000

[🌲 Evergreen] FOMO is not a strategy

Self-updating Profile README

Sun, 03 Nov 2024 00:00:00 -0000

[🌲 Evergreen] I am going to try to build a version of this for myself.

Week Notes 2024#5

Sun, 03 Nov 2024 00:00:00 -0000

[🌲 Evergreen] It has been a little over a month, time got away from me.