I would love to just say that if you are using claude code, you should no be on pro. I feel like all the people complaining are complaining that an agent cant handle the work of a developer for $20/m. Get on at least max 5, its a world of a difference.
Typing is not fun. It robs me of my craft of holding my pencil and feeling it press against the paper with my hand... LLMs are merely a tool to achieve a similar end result. The different aspects of software development are an art. But even with LLMS, I critique and care about the code just as much as if I were writing it line by line myself. I have had more FUN being able to get all of my ideas on paper with LLMs than I have had over years of banging my head against a keyboard going down the rabbit hole on production bugs.
On OpenAPI vs GraphQL: I disagree with the premise that OpenAPI achieves the same thing. GraphQL is necessarily tightly coupled to your backend — you can't design a schema that does something other than what's actually implemented. OpenAPI, on the other hand... I've seen countless implementors get it wrong. Specs drift from reality, documentation lies, and you're trusting convention. Sure, OpenAPI can do whatever you want, but for those of us who prefer convention over configuration, GraphQL's enforced contract is the whole point.
On authentication concerns: Yes, auth in GraphQL has varied implementations with no open standard. But REST doesn't thrive here either... it's all bespoke. This is a tooling problem, not a GraphQL problem. Resolvers become your authorization boundary the same way endpoints with controller actions do in REST. Different shape, same responsibility.
On type generation: In my experience, the codegen tooling with Apollo and Relay is incredible. I haven't seen anything on the OpenAPI side that comes close to that developer experience.
Hey HN – I built Han, a curated marketplace of Claude Code plugins that actually enforces code quality instead of just suggesting it.
The problem: developers constantly ship broken code because tools warn but don't block. Linters flag issues, but you can ignore them. Tests fail after pushing, not before.
Han solves this by running validation hooks at Claude Code "Stop" events – when you or a subagent finish working. If types fail, tests break, or linting errors exist, you can't continue until they're fixed.
Key features:
- 129 plugins covering 19+ languages, 30+ disciplines, 15+ testing frameworks
- AI auto-detection – analyzes your codebase and recommends the right plugins
- Smart caching – only runs checks when relevant files changed
- Zero config – MCP servers like GitHub integration work out-of-the-box with `gh auth token`
- Monorepo-friendly – automatically finds all applicable directories
The architecture separates techniques (jutsu – language/tool skills) from disciplines (dō – developer practices), reflecting that your identity as a frontend developer doesn't change when you switch from React to Vue.
Built with TypeScript, Ink (React for CLI), Commander.js, and the Claude Agent SDK. MIT licensed.