What Is an Agent Harness? How Coding Agents Actually Work
Inside Claude Code, Codex, Amp, and Factory: agent loops, tools, context, sandboxes, permissions, memory, and feedback.
Read about my deep dives and experiments with LLMs, AI Agents, and building with AI APIs.
Showing 12 of 57 posts
Inside Claude Code, Codex, Amp, and Factory: agent loops, tools, context, sandboxes, permissions, memory, and feedback.
Kimi K3 brings current frontier-level intelligence to open weights. Here is why that matters, how the architecture works, and every practical way to use it.
Build a Claude Cowork personal AI assistant that organizes files, creates documents, connects to tools, and runs recurring tasks without code.
Learn how Claude Managed Agents provides hosted containers, tools, sessions, and multi-agent orchestration for building autonomous agents.
See how Claude Code's agent loop handles prompts, tools, context, and verification, and apply the same design to your own agent harness.
I built a Claude Code skill that drafts and publishes my weekly newsletter to Kit, so it actually gets done every week.
Learn what vibe coding is, how it works, which tools to use, where it breaks, and the practices that make AI-generated software reliable.
Deploy governed Claude Agent Skills across a team. Learn distribution, access control, testing, updates, security, and ownership at scale.
Stop using other people's MCP servers and build your own. A step-by-step Python tutorial that takes you from zero to a working MCP server in under 30 minutes.
Build a personal Jarvis AI assistant with Claude Code that plans your day, manages projects, processes email, and remembers how you work.
How we turned Lenny's podcast and newsletter archive into 76 verified, installable product skills for Claude Code and other AI agents.
A practical guide to the AI coding tools, deployment platforms, databases, and payment systems you need to build and ship software quickly.
Learn the Ralph Wiggum coding-agent loop: a simple long-running workflow that uses repeated prompts and verification to ship working code overnight.
I ditched WordPress for Astro + Markdown. Now my blog is code, deployments are instant, and I've completely transformed my content creation process.
Built an AI executive assistant that runs entirely on your laptop—zero API costs, zero data leaving your machine, complete privacy.
Gemini 3 isn't just better—it's operating on a different level. Built a boxing coach, finance app, and city builder in real-time.
Automate VC due diligence: paste a startup URL, get competitive analysis, whitespace identification, and investment-ready insights in minutes.
Convert any blog post into a podcast with natural voice, emotional range, and realistic pacing—all powered by Cartesia's ultra-realistic TTS.
Build reusable Claude Agent Skills with SKILL.md, scripts, references, invocation controls, and practical examples for Claude and Claude Code.
Monitor competitor ads in real-time using browser automation that looks human, bypasses detection, and feeds AI-powered competitive intelligence.
Learn Factory Droid setup, Spec Mode, Missions, custom agents, skills, MCP, worktrees, model selection, and headless automation.
Stop manually checking competitor websites. AI-powered scraping that extracts clean data, detects changes, and generates actionable business intelligence.
700 million people, 18 billion messages weekly. Here's what they're actually doing—and why 73% of usage is now personal, not work.
Forget learning a new tool—master the universal patterns that work across Claude Code, Amp, Cursor, and any AI coding agent you'll ever use.
Learn Amp Code installation, capability modes, subagents, skills, plugins, MCP, shared threads, orbs, and remote agents in one practical tutorial.
Roll your own research agent that beats OpenAI's Deep Research for custom workflows—auditable, private, and integrated into your actual tools.
Build a working coding agent in Python with the Claude API. Add tool use, file editing, code execution, context management, and verification.
Forget static videos. Genie 3 generates entire interactive 3D worlds you navigate in real-time—the Will Smith spaghetti video just became ancient history.
Context beats prompts. Learn the systems that separate high-performing AI agents from the ones that hallucinate and fail—with real architecture patterns.
Learn Claude Code from setup to production workflows. Build a real app while mastering planning, agents, skills, MCP, hooks, testing, and deployment.
ChatGPT makes you faster, but it can make you dumber. Learn the active mode thinking strategy that keeps your brain sharp while using AI.
Generate high-performing content in hours instead of days by combining your unique perspective with AI's research, structure, and editing power.
Build 50+ production-ready automations without code—from lead qualification to content repurposing to customer support—using Make.com and AI.
Learn Cursor AI from first prompt to reviewed code. Use Agent mode, rules, context, background agents, the CLI, MCP, and safe review workflows.
Learn four production AI agent architectures, the components each needs, and practical patterns for designing agents that work reliably.
Ship MVPs in days by vibe coding with AI. Transform your ideas into working SaaS products using Cursor, Lovable, and iterative AI collaboration.
ChatGPT has become a digital yes-man that won't challenge bad ideas. Why flattery-optimized AI is dangerous for your thinking—and what to do about it.
o3 combines reasoning with tool use in ways Gemini can't match. See how it dominates research, vision, coding, and multi-step data tasks.
Learn Google ADK with practical Python examples for tools, multi-agent systems, workflow graphs, memory, evaluation, A2A, MCP, and production deployment.
Your content won't rank on ChatGPT if you don't optimize for it. The GEO playbook for getting your brand mentioned in AI recommendations.
Code by describing what you want in plain English. The mindset shift that lets you ship projects without memorizing syntax or debugging alone.
Build custom SaaS apps in minutes instead of months—no engineers required. How on-demand software is disrupting the billion-dollar SaaS industry.
Learn how to use Claude as a collaborator for content, code, and complex decisions—and get better results by letting it challenge your thinking.
Build production AI agents with the OpenAI Agents SDK. A practical Python guide to tools, handoffs, guardrails, sessions, human approval, and tracing.
Stop writing blog posts and publishing them manually. Watch how MCP enables Claude to research, write, and deploy content end-to-end without leaving your chat.
MCP isn't just another API wrapper—it's the protocol that turns your AI assistant from a chatbot into a digital worker with access to your entire digital world.
The universal translator between AI and your digital life. Learn how MCP gives Claude real power to act, not just chat, in 10 minutes.
OpenAI's most expensive model ever delivers... modest improvements. We tested it against Claude, Grok, and Gemini to see if the 15X price premium is worth it.
Build web-aware AI agents with Exa search, with practical code examples for research, competitive analysis, retrieval, and source selection.
Learn which SaaS ad campaigns to run at each stage of the buying journey and how to use AI to improve creative testing and performance.
Scaling parameters hit a wall, but reasoning models cracked the code—they think step-by-step like humans, solving problems once thought impossible for AI.
An AI agent that transforms your VC's private investment memos into polished blog posts automatically—with code you can clone and customize for your fund.
Tiny AI models running offline on your phone beat massive cloud models—here's why the future of AI is local, private, and completely free.
See exactly which AI tools a consultant uses for planning, writing, coding, and research—plus why they switch between Claude, ChatGPT, and reasoning models.
Bidding on competitor names isn't trash talk—it's the highest-intent audience ready to switch. Master the subtle art of stealing their customers profitably.
Forget viral growth hacks. This systematic process of personas, journeys, testing, and scaling has driven actual growth at dozens of startups.
When the market freezes, most teams cut marketing—but this playbook shows how obvious and non-obvious products can thrive by finding their new audience.
No posts found with this tag.