Posts

An open source agent harness is the runtime layer that turns a raw LLM into a working agent, handling the execution loop, tool calls, sandboxing, and context management so runs are reproducible. TrueForge is a TypeScript, MIT-licensed harness that delivers this in one process with SQLite locally or Postgres and Redis when hosted, claiming the same accuracy as Claude Managed Agents and deepagents at lower cost. What Is an Agent Harness and Why Does It Matter? Building an agent is easy; running one well is hard. The gap between a prompt that calls a model and a system that reliably completes multi-step tasks is exactly what an agent harness fills. A harness is the software layer that owns the agent execution loop: it decides when to call the model, how to route tool invocations, how to manage the growing context window, and how to keep the whole run safe and auditable. ...

Autolith is a Common Lisp terminal programming agent that runs inside a live SBCL image rather than wrapping another agent process. Instead of editing files and waiting for a rebuild, it can inspect, test, and modify its own running Lisp runtime, with changes taking effect immediately and recorded in an append-only mutation journal. It is a genuinely different paradigm from file-based coding agents like Cursor or Claude Code. What is Autolith? A Common Lisp agent with a live runtime Autolith, built by Lambda Symbolics, is a self-modifiable, general-purpose Lisp AI agent that executes inside a live Steel Bank Common Lisp (SBCL) image. The binary release carries its own SBCL 2.6.6, so you are not installing a wrapper that shells out to a separate agent process — the agent is the Lisp image, and the Lisp image is the agent. The prompt input is the REPL itself, and the harness can inspect and extend the very runtime it runs on. ...

WebMux is a self-hosted, browser-based remote workspace that tiles persistent SSH, mosh, VNC, and RDP sessions into a single scrollable web interface — effectively “tmux, but for the browser.” It started as a shared jump-box wall of terminals and has grown into a full multi-session terminal and desktop workspace with multi-user accounts, auto-reconnect, and OS service integration. What Is WebMux? WebMux is an open-source TypeScript application that turns your browser into a persistent remote workspace. Instead of running a single terminal in a web page the way tools like ttyd or gotty do, WebMux gives you a 2D tiled grid of sessions — SSH terminals, mosh connections, and even VNC or RDP desktop views — all living in one interface that survives browser closes and server reboots. ...

Multi GPU local inference means spreading a single large language model across two or more GPUs in your home server so you can run models that exceed the VRAM of any one card, or serve them faster. You do this by “drifting” the model’s layers or tensors across your GPUs using either pipeline (layer) parallelism or tensor parallelism, configured through tools like llama.cpp and vLLM. The right strategy depends on your interconnect speed, your hardware mix, and whether your model architecture even supports the split. ...

Code as an artifact means treating the source you commit as a regenerable output of a process — the prompts, plans, and agent context that produced it — rather than the durable end product of software development. When AI agents generate most of your code, the spec and the conversation become the real artifact, and versioning must capture both the code and the process behind it. This guide explains why traditional Git falls short, how to attribute every line to AI or human, and which workflows and tools keep AI-generated code trustworthy at scale. ...

Coding agent experience management is the practice of turning the session history your local coding agents generate into a searchable, migratable, and reusable asset instead of leaving it as scattered log files. Open-source tools such as AgentHist bring sessions from Codex, Claude Code, OpenCode, and Pi into one library for browsing, searching, exporting, and selective importing — including cross-machine migration and cross-agent conversion. AgentHist is a TypeScript CLI, MIT-licensed, supporting Linux, macOS, and Windows, created August 2026 (source: GitHub API). ...

Portable Handoff: How to Keep Context Between AI Chats, Models, and Coding Agents AI coding agents start every session with a fresh context window, so anything you do not persist is re-learned from scratch. A portable handoff file solves this: you write your instructions, preferences, and project conventions once, then symlink or import that single source of truth into every agent you use — Claude, Gemini, Cursor, or Codex — so context follows you across chats, models, and tools without drifting. ...

Coding agents lose context across repositories because a session is scoped to one repo while the system under investigation spans many. WCTX is a local-first MCP server that solves this by importing finished coding-agent sessions into a SQLite + FTS5 store, linking repositories with typed relations, and exposing seven tools that let a fresh session search prior evidence with git-based freshness verdicts. No cloud account, no embeddings, no transcript upload. ...

An LLM cost latency router is a decision layer that sends each request to the cheapest model that still meets your quality and latency bar, instead of calling one expensive model for everything. Real-time routing measures live cost and speed per model and picks the best option per request, cutting inference spend by 35-85% while retaining about 95% of top-model quality. This guide explains how it works, the strategies, and the numbers. ...

The Rapid-MLX DSH Provider is a native npm plugin for the DeepSeek Harness (dsh) that lets the agent read its model configuration straight from a running Rapid-MLX server’s /v1/models endpoint instead of hand-written settings.yaml. Instead of copying context windows, reasoning efforts, and max_model_len values by hand — and re-copying them every time you switch models — the provider serves those facts directly from Rapid-MLX’s JSON schema, keeps reasoning control truthful, and compacts the context at the real memory limit of your Apple Silicon machine. ...