# The Null Hypothesis

> Front-end, LLM tooling, and what actually works.

## Posts
- [Adding llms.txt to Your Astro Blog](https://kumak.dev/llms/adding-llms-txt-to-astro.txt): How to implement the llms.txt standard in Astro without dependencies. Three endpoints, ~150 lines of TypeScript, and your content becomes agent-accessible.
- [Testing in the Age of AI Agents](https://kumak.dev/llms/testing-in-the-age-of-ai-agents.txt): When code changes at the speed of thought, tests become less about verification and more about defining what should remain stable.
- [Self-Documenting CLI Design for LLMs](https://kumak.dev/llms/self-documenting-cli-design-for-llms.txt): Agents start fresh every session. Instead of dumping docs upfront, build tools they can query. One take on agent-friendly tooling.
- [CLI vs MCP on Chrome DevTools Protocol](https://kumak.dev/llms/cli-vs-mcp-benchmarking-browser-automation.txt): I ran benchmarks comparing CLI tools against MCP servers for browser automation. 13x more token efficient with CLI. Here's what I found.
- [How My Agent Learned GitLab](https://kumak.dev/llms/how-my-agent-learned-gitlab.txt): Teaching an agent to use CLI tools isn't about writing perfect documentation. It's about creating a feedback loop where the tool teaches, the agent learns, and reflection builds institutional knowledge.
- [When CLI Tools Are Enough](https://kumak.dev/llms/when-cli-tools-are-enough.txt): One year after MCP's launch, Anthropic published a solution to context overhead: filesystem-based tool discovery. The pattern looked familiar.

## Optional
- [About](https://kumak.dev/about): About the author
- [RSS Feed](https://kumak.dev/rss.xml): Subscribe to updates
- [Full Content](https://kumak.dev/llms-full.txt): Complete post content for deeper context