LLM-Wiki v0.16.0 introduces Query Lite, a compact read-only query protocol that reduces Claude Code wiki query instructions by 72.54%, alongside token benchmarks and reproducible benchmark lanes for C…
Dax Raad, co-founder of OpenCode, revealed on the Syntax podcast that his team's AI token usage increased 5x after gaining early access to an unreleased AI model, and when access ended, team members m…
New testing by AI consultancy Systima found that agent harness configuration can significantly inflate token overhead and enterprise AI deployment costs, with research showing that changing the harnes…
Harness Engineering built a fully automated coding loop inside OpenCode using DeepSeek-V4-Pro and DeepSeek-V4-Flash, enabling low-cost agentic code iteration. The loop, inspired by Andrew Ng's breakdo…
George Hotz, founder of tinygrad and comma.ai, published a July 12th blog post arguing that AI doom rhetoric is being used to justify high valuations for frontier AI labs, while the actual progress of…
A developer rebuilt their agentic coding kit, Agentic Coding Kit v6, to eliminate wasteful process loops that made small tasks expensive. The new version enforces a leaner rule where the main coding s…
A developer built Bothread, a free, open-source local server that coordinates multiple AI coding agents via the Model Context Protocol to prevent file overwrites. In a demo, three agents from differen…
Open-source coding agents can now run entirely on local consumer hardware such as a 2020-era NVIDIA RTX 3090, eliminating usage-based pricing from AI vendors. The setup uses a quantized Gemma4 model s…
A developer placed a logging proxy between two coding agents and their model to measure token overhead. Claude Code shipped roughly 33,000 tokens of scaffolding before a 22-character prompt, while Ope…
A developer released skill-auditor v0.1.0, an agent-agnostic CLI tool that scans AI agent skill collections for structural integrity, broken references, and hardcoded secrets. In a real audit of 200 H…
A developer who builds a Paseo competitor reviewed Paseo v0.1.107 from the terminal, finding that saying 'hi' cost $0.29 and that every run creates a fresh workspace. The reviewer praised the daemon's…
A Hacker News comparison on 2026-07-13 revealed a 26,000-token gap between coding agents Claude Code (33,000 tokens overhead) and OpenCode (7,000 tokens overhead), shifting focus from model capability…
A Hacker News user asks for recommendations on securely containerizing AI agent harnesses to prevent unauthorized file system access, citing their use of OpenCode with DeepSeek on Linux Mint Debian Ed…
Developer saadev0 has published a portable setup for OpenCode using the opencode-antigravity-auth plugin, enabling Google OAuth-based access to Antigravity quota for Gemini and Claude models. The conf…
A wire-level analysis by Systima reveals that Anthropic's Claude Code sends approximately 32,800 tokens before reading a user's prompt, compared to 6,900 tokens for OpenCode—a 4.7x overhead gap. The s…
A wire-level analysis by Systima published this week found that Claude Code sends approximately 33,000 tokens to Anthropic's API before processing a single word of your prompt — 4.7 times more than Op…
A developer benchmarked GPT-5.6 against Claude Code and OpenCode, finding GPT-5.6 delivers a 2.2x speed advantage over OpenCode and reduces costs by 21% compared to Claude Code. The tests measured spe…
Developer sin-ack released Agent-run, a tiny standalone binary that runs coding agents like Pi, OpenCode, Codex, and Claude in a sandboxed Linux environment using bwrap. The tool, available on GitHub,…
Claude Code consumes approximately 33,000 tokens in system prompts, tool schemas, and scaffolding on its first request, compared to OpenCode's 7,000 tokens, a 4.7× overhead that increases session cost…
A developer built Kote, a tool that automatically captures engineering context from AI chats, Git pushes, and voice notes to prevent knowledge loss, and made it self-hostable with support for custom L…