horn.call()

> Where music, humanity, and technology intersect

Recent Posts

When Your Agent Burns 71% of Your Weekly Tokens Overnight

πŸ“… February 16, 2026 πŸ“ Technical 🏷 ralph-docker , autonomous-agents , token-management , observability , claude-code , post-mortem

A Ralph Docker session ran overnight with unlimited iterations, creating 67 empty workspace branches and consuming 71% of my weekly Sonnet token allotment β€” with zero productive output. Here's the forensic analysis and the fix.

Read More β†’

Making Ralph Docker Plug-and-Play

πŸ“… February 12, 2026 πŸ“ Technical 🏷 ralph-docker , claude-code , autonomous-agents , developer-experience , observability

Ralph-docker now has a built-in setup command that interviews you and generates all project files via Claude β€” no external skill required. Plus, a rewrite of the README to put observability front and center.

Read More β†’

Chasing Overflow Down the CSS Rabbit Hole

πŸ“… February 9, 2026 πŸ“ Technical 🏷 css , hugo , responsive-design , debugging , claude-code

A debugging session where fixing one CSS overflow led to discovering Hugo's hidden table-based syntax highlighting, duplicate asset files, and the joys of browser caching.

Read More β†’

Reviewing and Fixing Four PRs in One Session

πŸ“… February 9, 2026 πŸ“ Technical 🏷 go , code-review , rabbitmq , concurrency , testing , claude-code

A session where Claude Code reviewed four open PRs on a Go/RabbitMQ event system, identified bugs ranging from mutex contention to JSON schema mismatches, and fixed them all β€” merging three and leaving one ready to go.

Read More β†’

Building an AI Agent Orchestrator from Scratch in Python

πŸ“… February 7, 2026 πŸ“ Technical 🏷 python , agents , design-patterns , oop , testing

Building a multi-agent system incrementally across 4 assignments β€” agent loops, pluggable tools, bounded memory with escalation, and orchestration β€” all in pure Python with 50 tests and zero LLM dependencies.

Read More β†’

Implementing calloc: Why Two Parameters Matter for Security

πŸ“… December 13, 2025 πŸ“ Technical 🏷 c , memory-management , malloc , security , overflow-protection

Building calloc for a custom malloc implementation and discovering why its two-parameter design isn't just conventionβ€”it's a critical security feature that prevents integer overflow attacks.

Read More β†’

Meta: Improving the Development Workflow Itself

πŸ“… December 13, 2025 πŸ“ Meta 🏷 meta , workflow , automation , claude-code , tooling

Taking a step back from coding to improve the tools that document the coding processβ€”fixing the /blog and /resume slash commands to be more robust and token-efficient.

Read More β†’

Improving the Blog Workflow: Independent Perspectives

πŸ“… November 22, 2025 πŸ“ Technical 🏷 meta , documentation , ai , claude , workflow , process

Refining the /blog command workflow to ensure Claude's technical analysis remains independent of user reflections, preventing hallucination and maintaining separate perspectives.

Read More β†’