Tag
Benzi is a compiler-backed coding agent that first compiles a codebase into a queryable map of calls, data flow, and class hierarchies, reportedly hitting 78.2% on SWE-bench Verified with DeepSeek V4 flash at under 10 cents per fix. It claims 2x less source code read than Claude Code while being 40% faster and 2x cheaper, and ships as a VS Code extension, headless runtime, and MCP tools.
Benzi is an AI coding agent that outperforms competitors on benchmarks by reading less source code and using deterministic tool calls, achieving high efficiency in code intelligence tasks.
Benzi is a code intelligence harness that benchmarks show it outperforms Claude Code and CodeGraph, resolving 78.2% of real GitHub issues at under 10¢ per fix.
The author developed 'mex,' a tool that maintains persistent memory for coding agents by validating and repairing the memory against codebase changes, ensuring it remains trustworthy over time.
Zijian Wang announces hosting the 5th DL4Code workshop at ICML this Friday with an impressive speaker lineup and mentions two past interns presenting their papers at the main conference.
DeusData has open-sourced an MCP server called codebase-memory-mcp, which pre-parses code repositories into persistent knowledge graphs, providing AI coding assistants with sub-millisecond code structure query capabilities. It supports 158 languages, reduces token consumption by approximately 99% compared to traditional grep, and has extremely fast indexing speed.
This article details the latest progress of GLM-5.2 in Agentic RL, including the introduction of slime infrastructure, shifting from GRPO to PPO for handling long trajectories, and an online anti-cheat mechanism; it also explores Qwen's research on verifier quality, proposing three dimensions of scalability, faithfulness, and robustness, and designs multiple verification strategies for different tasks to improve the reliability of reward signals.
codebase-memory-mcp is a tool written in pure C that parses the entire codebase into a knowledge graph, supports 158 programming languages, is compatible with 11 AI coding agent tools, greatly improving AI's understanding of project structure and reducing token consumption.
This survey paper systematically reviews multimodal code intelligence systems that generate and reason with code from visual inputs like screenshots and charts, categorizing approaches across GUI, scientific visualization, structured graphics, and emerging frameworks while proposing verification-centered future research directions.
Discussed how GitHub Copilot CLI integrates language server (LSP) to obtain semantic information, thereby enhancing code understanding in AI programming, and points out that AI programming needs to evolve from the text search layer to the semantic layer and runtime layer.
Repowise is an open-source tool that indexes codebases into five intelligence layers—dependency graph, git history, auto-generated docs, architectural decisions, and code health—and exposes them to AI coding agents via MCP tools for more accurate context and fewer tool calls.
Repowise is an open-source tool that indexes codebases into four intelligence layers (dependency graph, git history, auto-documentation, architectural decisions) and exposes them via seven MCP tools to AI coding agents like Claude Code, achieving up to 27× token savings while maintaining answer quality.
GitNexus is an open-source knowledge graph engine that indexes codebases, maps dependencies and execution flows, and integrates with AI coding assistants via MCP, providing full architectural clarity to prevent breaking changes.
Codebase-memory-mcp is an ultra-fast code intelligence engine for AI coding agents that indexes entire repositories in milliseconds and answers structural queries in under 1ms using tree-sitter AST analysis and a persistent knowledge graph, with support for 158 languages and 14 MCP tools.
CodeGraph is an open-source tool that creates a pre-indexed knowledge graph of a codebase, enabling Claude Code's exploration agents to query symbol relationships and call graphs instantly, reducing tool calls by up to 96% and exploration time by 77%.