A new open-source project called Whetstone hit Hacker News on August 9, offering developers something refreshingly honest in the AI tooling space: twenty Claude Code skills, each one painfully extracted from actual production failures rather than theoretical use cases.

The Philosophy Behind the Project

The project's approach flips the typical tutorial narrative. Instead of showcasing best-case scenarios and ideal prompt patterns, Whetstone documents what went wrong first. Each skill entry starts with a real failure mode—a context window overflow that corrupted a migration script, a model hallucination that created non-existent API endpoints—then walks through how Claude Code eventually handled it correctly after the developer learned to structure their prompts differently. The site is hosted at whetstone.akbarsha.dev and appears designed as both a learning resource and a practical reference library for developers already working with Anthropic's coding agent in real-world development environments.

Why This Matters for AI Agent Workflows

Claude Code adoption has accelerated through 2026, but documentation around production pitfalls remains scattered across Discord servers and GitHub issues. Whetstone attempts to consolidate these lessons into something more structured—skills that developers can internalize before hitting the same walls themselves. The project's author frames it as "distilled from one real failure" per skill, suggesting each entry represents a complete debugging narrative rather than abstract guidance. This case-study approach resonates with how senior engineers actually learn: through stories of what broke and why.

Community Reception

The Show HN post gathered modest early engagement on August 9, with developers weighing in on which failure modes resonated most with their own experiences using AI coding assistants in complex codebases.

Key Takeaways

  • Whetstone prioritizes learning from production failures over polished success stories
  • The library targets Claude Code users working in real development environments
  • Open-source structure allows community contributions of additional failure-derived skills

The Bottom Line

In the crowded AI tooling space, raw honesty about what breaks beats curated demos every time. Whetstone earns attention by doing something most projects avoid—leading with failure.