herdr vs cmux, after two months on each
After two months on each, developer herdr 0.7.4 and cmux 0.64.x differ mainly in architecture: herdr runs on the server and attaches via SSH, while cmux is a macOS app with a new SSH daemon. The autho…
After two months on each, developer herdr 0.7.4 and cmux 0.64.x differ mainly in architecture: herdr runs on the server and attaches via SSH, while cmux is a macOS app with a new SSH daemon. The autho…
Developer Simon Willison runs coding agents Pi and Claude Code on a MacBook Pro with permissions on and on disposable Dell OptiPlex micros with scoped credentials and no personal keys, arguing that th…
MCP servers should expose separate tools with precise names and descriptions rather than a single generic tool, according to a guide on designing MCP tools for agents. The approach helps models choose…
Pydantic AI structured outputs and evals now support Amazon Bedrock, providing decoder-level guarantees for LLM output shape and content validation. The framework compiles a grammar from a Pydantic Ba…
FastMCP, a framework for building MCP servers, faces security and access control challenges when connecting to real company data. Developers must address authentication, logging, and tool selection to…
A debate in AI agent development argues that MCP servers are unnecessary for single-user tasks, where simple 'skills'—folders of instructions and scripts—suffice. Skills run in the user's own context …
A developer introduces a method to anchor software specifications to code using ast-grep structural rules instead of file paths and line numbers, preventing spec drift during refactors. The approach u…
Employment for software engineers aged 22 to 25 has fallen roughly 20% since late 2022, according to Stanford analysis of 25 million payroll records, as AI tools and a broader recession squeeze entry-…
AI coding tools are reshaping software engineering, shifting focus from writing code to evaluating what to build and judging output quality. Anthropic's Mythos model found thousands of bugs but patche…
Open-weight local language models in mid-2026 have nearly matched frontier performance for everyday tasks, driven by engineering advances in sparse attention and mixture-of-experts architectures. Deep…
Local large language models have nearly caught up to frontier models for everyday tasks as of mid-2026, driven by engineering advances in sparse attention and mixture-of-experts architectures that red…
Developer Simon Willison reports that herdr, a Rust-based terminal multiplexer, effectively replaces tmux for managing multiple parallel AI agents on remote servers by providing a sidebar that shows e…
Developer Matt Coles built lgtmaybe, an open-source PR reviewer that works with OpenAI, Anthropic, OpenRouter, Bedrock, Vertex, or local Ollama models. The tool posts inline comments and summaries on …
Developer Matt Coles has built lgtmaybe, a provider-agnostic PR reviewer supporting six model backends with a single --provider flag, shipping as a PyPI CLI and GitHub Action. He also maintains a home…
Paul Kukiel and Melchi Salins are recommended as valuable resources for their expertise in AWS, Python, AI-assisted coding, and AI infrastructure, with Melchi's KV cache explainer featuring an interac…