Loop engineering is a deterministic control loop around an AI coding agent — schedule, durable state, and verification — that lets developers move up from prompting to designing the machine. Boris Che…
A developer created a Python tool to export shared DeepSeek conversations to Markdown, JSON, or plain text. The script automatically bootstraps its environment using uv, the fast Python package instal…
A developer highlights the Kesselman List of Estimative Words, a calibrated set of probability terms for intelligence analysis, as useful for large language models. The list assigns specific percentil…
A developer known as @99hud has published a guide on building a personal AI assistant similar to JARVIS using Claude and MCPs. The system automates daily briefings by pulling metrics, revenue, ad data…
Poolside's Laguna S 2.1, a 118B MoE coding model, runs on a single RTX 5090 (32 GB) at ~19 tok/s decode and ~60 tok/s prefill using auto-fit layer placement. A developer achieved this by packing full …
A developer has created agent_cli, a cross-agent command-line tool that unifies five AI coding agents—Claude, Codex, Cursor, Copilot, and Grok—behind a single interface. The script supports both read-…
A developer shared a workflow using Claude to automate job applications, where the AI analyzes a resume, finds matching LinkedIn postings, builds a compatibility spreadsheet, and submits applications.…
An engineer shared a Python script, agent_expanded.py, that implements a simple agent loop for a large language model with a 1,050,000-token context window. The script uses a custom shell tool to exec…
A developer created claude-hist, a command-line tool that uses fzf to fuzzy-find and resume past Claude Code sessions across all projects. The tool searches session titles, git branches, and conversat…
A benchmark comparing LLM Gateway and OpenRouter for Claude Haiku 4.5 found LLM Gateway is ~35% faster to first token cold (906 vs 1392ms) and ~34% faster warm (814 vs 1232ms) at the median. The TTFB …
A developer achieved ~28 tok/s single-stream decode of DeepSeek-V4-Flash (284B MoE) on a single RTX 5090 with dual Xeon Gold 6138 CPUs using a CPU-offloaded expert recipe. The setup uses ktransformers…
An engineer created a minimal AI agent in just 9 lines of Python code, using a custom tool to execute shell commands and a loop that processes user input through a language model. The agent, which cal…
A developer created a PowerShell script to recover Claude Code Desktop sessions that are missing from the sidebar on Windows. The script scans ~/.claude/projects/ for .jsonl files and generates local_…
A developer argues that major AI companies are deliberately suppressing advanced sampling techniques like min-p, mirostat, and dynamic temperature, offering users only basic controls while hobbyist to…
A developer has created a Claude-powered household meal planner called Meal Prep that runs locally via the Claude Code desktop app. The system interviews users about dietary preferences, allergies, ki…
A developer created a transparent theme for OpenCode that allows Ghostty terminal transparency to show through, featuring a cyan/slate accent palette.…
A developer reports that Flatpak apps disappear in Elementary OS 8.1 after certain package installations or repository additions. The developer shares a troubleshooting solution involving reinstalling…
A developer created a runtime tool that enables the hidden 'Control other devices' feature in Codex Desktop for Windows without modifying the installed binary. The tool overrides a Statsig gate and sp…
A developer has documented a configuration for Herdr, a persistent AI agent workspace tool for macOS, that enables reboot-safe layout and session resume for Claude Code and OpenCode. The setup require…
A developer has published a generic architecture blueprint for building a persistent, always-on personal AI system that acts as a second brain for any LLM. The system is designed to be brand- and fron…