Viktor Soroka·Oct 18I Gave Claude Code an LSP. Then I Counted Whether It Mattered.Sessions used it when asked who calls a function and skipped it mid-task. Nudges changed nothing, grep never missed a caller in 288…A response icon1A response icon1
Harshana·Oct 9What 16 Papers Say About Language Server Performance (And What They Don’t)A quick note first: this is not a peer-reviewed paper. It’s a literature review I wrote for my independent study at the University of…
Harshana·Sep 28Your WASM Function Call Isn’t Async. It’s a Full Stop.While rebuilding WSO2’s Micro Integrator language server in TypeScript, I compiled Apache Xerces-C++ to WebAssembly so we could do real XSD…
Srilalitha Vishnubhotla·Aug 7Inside Claude Code: How Tree-sitter, LSP, and MCP Kill Token InflationHow Anthropic decouples structural code understanding from the cloud to deliver fast, hyper-efficient local engineering loops.
Harshana·Aug 4How Language Servers Actually Work Under the HoodBefore this year, I used language servers every single day without knowing they existed. Red squiggles under broken code, autocomplete…
Fernando Paladini·Jul 25Stop Grepping Symbol Names: I Shipped Locus, a Lightweight MCP Server for AI Coding AgentsTL;DR — AI agents are great at editing files, but grep is the wrong tool for definitions, references, and types. Locus is an open-source…
Vignaraj Ravi·Jul 5Why LSP Is the Most Underrated Superpower in AI-Assisted DevelopmentThere is a quiet gap between what AI coding agents claim to do and what they actually understand about your code. They read files. They…
Satish Krishna·Jul 1Don’t RAG Your CodebaseFor navigating code, a language server is right where a code graph is wrong, and it never goes stale.
Arunvenmany·Jun 13Beyond the Pop-up Menu: The Reality of Implementing Inline Completions in Language ServersIf you’ve spent any time inside a modern IDE lately, you’ve experienced Inline Completion. The days of the passive, token-by-token…
Anton Babenko·May 18One model is a guess. Three that agree is a plan.Why I shipped multi-model consensus as a plugin, plus two quieter tools that keep agents honest.