engineering
The Command You Approve Is Not the Command That Runs →
Why approved commands run as something else: shell expansion rewrites what your agent's safety check read. A case for guarding at execution.
Thoughts on AI, engineering, and building the future.
Latest
View All → engineering
Wax: One File That Survives Hard Kills and Context Resets
I built Wax, a single-file crash-safe memory engine for AI agents on Apple Silicon: WAL durability, hybrid RAG, bitemporal facts, no cloud dependency.
engineering
Inside Swarm's WebSearchTool: How I Built It to Fit a 4K Context Window
I built Swarm's Swift web tool for 4,096-token on-device LLMs. Here's the tiered envelope, the Wax-without-embeddings bet, and fourteen decisions behind it.
engineering
How I Conducted Deep Research with a 3B On-Device Model and a 4K Context Window
A 3B on-device model with a 4K context window produced a 2,336-word grounded research report using live web search. Here's the architecture that made it work.
engineering
EdgeRunner Beats llama.cpp: A Swift/Metal LLM Engine on Apple Silicon
Built EdgeRunner in pure Swift and Metal from scratch over a weekend. On Qwen3-0.6B Q8_0, it hits 212 tok/s on M3 Max, beating llama.cpp by 16%. Here's what w
Explore Topics
Deep dives into specific areas