Live Notes from Inside the Vibe Coding Token Casino
Where every build gets you a dopamine high but you forget to check the balance.
Latest
What Is Cognitecture (And Why I Started This Blog)
Cognitecture means bounded delegation: AI agents act under independent checks, a named owner, clear limits, and a way back when things fail.
The Config-as-Code Multiplier: A Full Marketing Data Stack. Solo. In Weeks.
How I used config-as-code tools and AI agents to build a marketing data platform alone in six weeks.
What I'm Reading
Obsidian + Claude Code: Changes How You Use AI
The article discusses the integration of Obsidian, a knowledge management tool, and Claude AI to enhance productivity in note-taking and information processing. It highlights how this combination can streamline tasks by allowing users to leverage AI for summarization, research, and organization...
Evaluating AGENTS.md: Are Repository-Level Context Files Helpful for Coding Agents?
The article explores the effectiveness of using repository-level context files like AGENTS.md to customize coding agents in software development, which can be created through manual or automated methods.
“Across multiple coding agents and LLMs, we find that context...
The analysis across various coding agents and large language models (LLMs) indicates that providing context files decreases task success rates and simultaneously raises inference costs by more than 20% compared to scenarios without repository context.
The Context Development Lifecycle: Optimizing Context for AI Coding Agents
The article discusses a new lifecycle model focused on optimizing the 'context' for AI coding agents. This includes stages of generating, evaluating, distributing, and observing context data essential for AI's effective code generation.
I Spent 5 Days Debugging My OpenClaw Agent's Memory. Here's Everything I Learned.
The author details their five-day journey debugging an OpenClaw agent's memory issues, sharing lessons learned about troubleshooting AI agents, common pitfalls in agent development, and strategies for effective debugging.
My Latest Bookmarks
GitHub - ComposioHQ/agent-orchestrator: Open-source multi-agent orchestrator for Claude Code — manage parallel coding agents across projects
The Agent Orchestrator manages parallel AI coding agents across projects, enhancing productivity for knowledge workers through real-time monitoring and integration flexibility.
GitHub - blader/humanizer: Claude Code skill that removes signs of AI-generated writing from text
The Humanizer skill for Claude Code refines AI-generated text to sound more human-like by addressing common issues such as significance inflation and promotional language, enhancing the natural flow of writing.
GitHub - botiverse/agent-vault: Keep your secrets hidden from AI agents.
AGENT-VAULT safeguards secrets like API keys from exposure during service setup with AI agents by using placeholders and an encrypted local vault.
GitHub - yigitkonur/cli-continues: One command to resume any AI coding session. Codex • Claude • Copilot • Gemini • OpenCode
Seamlessly continue AI coding sessions across tools like Claude, Copilot, Gemini, Codex, OpenCode with `cli-continues`, enabling cross-tool handoff and automatic session discovery through a TUI.
dmux - Parallel agents with tmux and worktrees
Dmux streamlines running multiple AI coding agents in parallel via tmux and git worktrees, aiding knowledge workers in efficient branching, development, and merging.