From the Blog

Thoughts on Open Source, AI, web development, and building software

Found 112 posts

2026-09-26

How my browser game runs in VR on a Meta Quest

Bannermarch is a 3D game that runs in the browser. Now you can open it in the Meta Quest browser, press Enter VR and walk around inside it, with nothing to install. Here's how the browser does VR, how Babylon.js helps, and what went wrong along the way.

software-developmentgame-development

2026-09-26

10 min read

Signing in a VR headset the way your TV does

Typing a password on a Meta Quest means pecking at a floating keyboard, so Bannermarch signs headsets in like a TV does. The headset shows a code and you approve it at /pair on your phone. Here's how the pieces fit, why someone who reads the code can't use it, and the reverse flow I added because you can't see your phone with a headset on.

software-developmentsecurity+1

2026-06-18

4 min read

My agent kept losing facts it had already saved

AIdaemon could save a fact and then fail to find it when I asked for it. Semantic search ranks by what's on the same topic, so the short fact holding the real answer gets buried under wordier ones nearby. A reranker reads each candidate against the question and puts the right one back on top.

aisoftware-development+1

2026-06-08

13 min read

Local Gemma was too slow with AIdaemon until I fixed llama.cpp and the prompt size

I wanted AIdaemon on local Gemma 4 26B through llama.cpp, not Ollama. Generation ran at ~45 tok/s on an M4 Pro. Agent turns still felt stuck because prefill on 14k-token prompts took 8 to 9 seconds before the model wrote a single word.

aisoftware-development+1

2026-05-20

12 min read

How production-grade AI agents actually work

A plain-English walkthrough of the control plane, tool plane, memory, guardrails, sub-agents, and observability behind production-grade AI agents.

AIArchitecture+1

2026-05-15

3 min read

How to work with any generative AI

Switching from ChatGPT to Claude to Gemini to Grok, or to any open source model, shouldn't mean relearning how to prompt, what to trust, and what's safe to hand off. Under the hood they share the same handful of traits, and the same few habits get good results from all of them.

aifundamentals

2026-05-14

2 min read

Skills, CLAUDE.md, subagents, hooks, or MCP, when to use which

Claude Code has five customization layers. Pick the wrong one and you get a Skill that never fires, a hook doing subagent work, or a CLAUDE.md so bloated it eats half your context before you type. Here's where each one belongs.

aiclaude-code+1

2026-05-11

5 min read

Cutting OpenNext KV Storage Costs on Cloudflare Workers

OpenNext on Cloudflare stores ISR cache in KV, and every deploy quietly adds a new universe of keys that never get cleaned up. Here is how one of my sites went from $0.50/day in KV storage to under three cents.

cloudflarenextjs+1

2026-05-11

3 min read

Notes on installing Microsoft Clarity

I added Microsoft Clarity to one of my projects this week. It is free, the setup took ten minutes, and it has already told me more about how people use the site than Google Analytics ever has.

web-developmentanalytics+1

2026-05-03

5 min read

How to keep multiple coding agents from overwriting each other

The practical workflow I use for parallel AI coding agents: one worktree per agent, one branch per task, clear file ownership, protected main, and shared instructions through AGENTS.md or CLAUDE.md.

aisoftware-development+1

Get the latest posts and insights delivered to your inbox.

No spam. Unsubscribe at any time.

From the Blog

How my browser game runs in VR on a Meta Quest

Signing in a VR headset the way your TV does

My agent kept losing facts it had already saved

Local Gemma was too slow with AIdaemon until I fixed llama.cpp and the prompt size

How production-grade AI agents actually work

How to work with any generative AI

Skills, CLAUDE.md, subagents, hooks, or MCP, when to use which

Cutting OpenNext KV Storage Costs on Cloudflare Workers

Notes on installing Microsoft Clarity

How to keep multiple coding agents from overwriting each other

Stay Updated