What Google's defamation banner says about LAP
Reconstructing confidence windows from Google’s new defamation-removal banner suggests LAP Coffee’s chain-wide rating is closer to 3.8 than the displayed 4.6 — a full star of missing reviews.
Reconstructing confidence windows from Google’s new defamation-removal banner suggests LAP Coffee’s chain-wide rating is closer to 3.8 than the displayed 4.6 — a full star of missing reviews.
Using Claude Code on autopilot to build a bash AST parser in Rust, driven by Parable’s test suite. The result: rable, a 99.9%-accurate parser that outperforms tree-sitter-bash across 1,783 test cases.
Extending tokf’s token compression to work inside Make, Just, and git hooks using shell injection and PATH shimming to intercept nested subcommands.
tokF is a configurable CLI output filter built in Rust that uses TOML pipelines to strip, extract, and compress command output before it hits an LLM — achieving 90% token reduction in real-world use.
Building a cargo subcommand that detects duplicate Rust code by normalizing ASTs - replacing names with placeholders, erasing literals, and fingerprinting the structural skeleton.
A week of debugging OAuth flows, stateless handlers, and Claude.ai integration issues - ending with a pivot to Claude Code and 33% cost savings.
An experiment in building copy/paste MCPs for AI code refactoring, with tools to benchmark them. Hypothesis: reduce output tokens. Result: not yet.
From a single messy file to multi-agent pipelines and proper engineering. A year of iterating on AI-generated satirical content.
Frustration, a cheap domain, and Claude turned into an AI satire engine with 1197 posts and counting. Here’s how STFU works.