The AI Builder Newsletter Weekly deep-dives on Claude Code, MCP, AI agents, and building real products with AI. Written by AI Jason for ambitious builders. Join 5,000+ builders. Free forever. No spam. A full set of free AI builder courses. No signup, no paywall. These are the same workflows we use to ship real products with AI, written up so you can copy them step by step. Start with Claude Code basics, then go deep on MCP and building AI agents from scratch. Every lesson ships working code, not theory. Pick a track and start building. Claude Code Fundamentals Anthropic's terminal coding agent, end to end 1 Getting Started 2 Workflows & Techniques 1 Getting Started Video Everything You Need to Ship with Claude Code The complete Claude Code reference - install, configure, master. Install, Configure, and Ship Your First Feature Install Claude Code, run your first terminal session, understand the basics. The CLAUDE.md Guide: Configure for Any Project Project config, memory, custom instructions with CLAUDE.md. Karpathy's CLAUDE.md Rules Karpathy's real CLAUDE.md config - what he actually uses. .agent Folder: Context Engineering Context engineering with a structured .agent documentation system. Shared lesson with the AI Agents course. 2 Workflows & Techniques 10 Prompting Patterns That Actually Work Prompting patterns that consistently get better results. Agent Hooks: Rules Your AI Can't Ignore Lifecycle events, matchers, and 7 copy-paste hook recipes. Video SubAgent: Context Isolation 3x output with parallel task delegation and clean context. Video Agent Teams: Run an AI Project Group Team lead, teammates, shared tasks, and direct messaging. Video The Sidekick Paradigm: Orchestrator-Worker Agents Strong model plans, cheap model executes - cut AI coding cost ~80% without losing quality. Dynamic Workflows: Orchestrate Subagents at Scale Move the loop into a script - run up to 1,000 subagents without flooding context. Worktree: File Isolation for Parallel Agents Give every agent its own directory - no more write conflicts. TodoWrite vs Task: Agent Task Management Session checklists vs persistent task graphs - when to use which. Video /goals Command: Tips & Mistakes Goal-driven agent sessions - tips and mistakes. Shared lesson with the AI Coding Tools course. Run Parallel Sessions Like Karpathy Run parallel Claude Code sessions for maximum throughput. Cut API Costs 73% (7 Strategies) 7 strategies to cut API costs without sacrificing quality. Official Plugin Library + SKILL.md Official Anthropic plugin library and SKILL.md auto-loading. SKILL.md Auto-Loading (No Marketplace) Drop a SKILL.md and Claude Code auto-loads it. Video Anthropic's 300+ Skills Playbook Anthropic's internal playbook for building reliable Skills. Sam Altman's Skills Framework for Codex Most agent failures are context problems, not model problems. Video Connect MCP Servers to Claude Code Extend Claude Code with custom tools via MCP. Shared lesson with the AI Agents course. Build AI Agents From first principles to production 2 Build From Scratch 3 Tools & Infrastructure Video AI Agents: The Complete 2026 Roadmap Complete guide to building, deploying, and scaling AI agents. What Is an AI Agent? What an AI agent is - planning, memory, tools, action. Video Function Calling: How LLMs Use Tools The menu-order-serve mechanism behind every agent action. Give Your Agent Tools (Function Calling) Give your agent real tools - function calling and MCP. Agent Memory: Short-term and Long-term Agent memory - short-term, long-term, and retrieval. Video Memory Systems: Episodic, Semantic, Procedural The 3 memory types, storage layers, and MemGPT vs Mem0 vs Claude Code. Multi-Agent Systems: Orchestration Multi-agent systems - orchestration and delegation. Deploy Agents to Production Deploy agents to production - reliability and monitoring. Karpathy's Agentic Engineering Framework Karpathy's framework for serious AI builders. Karpathy's agents.md: 4 Rules What Karpathy's agents.md would look like - four rules for production agents. Software 3.0: Context Window = Program The paradigm shift from traditional software to AI-native code. Karpathy's LLM Wiki Pattern Build a personal knowledge base that compounds. 2 Build From Scratch Build an Agent in 60 Lines of Python Build an AI agent from scratch in Python - step by step. Multi-Agent Coordinator + Workers (200 lines) Multi-agent coordination in Python - real working code. Video Hermes: Self-Hosted, Never Forgets Self-hosted AI agent that retains memory across sessions. Gemma 4: Free Local Agents with Ollama Run free agentic AI locally with Ollama + Gemma 4. 3 Tools & Infrastructure Video MCP 101: Build Your First Server Build your first MCP server from zero. MCP Internals: Client-Server Communication STDIO vs SSE, JSON-RPC, and the full 6-step interaction loop. MCP Security: 6 Attack Vectors Tool poisoning, exfiltration, rug pulls - and the 5-step audit. Video Why MCP is Dead & Skills Replace It Where Skills replace MCP servers - and where they don't. Shared lesson with the Claude Code course. Video WebMCP: Agents on Websites How websites expose structured tools to in-browser agents. Context Engineering: The Real Bottleneck 4 strategies, 4 failure modes, and KV-cache economics. RAG vs Long Context vs Fine-Tuning When each wins, the modern pipeline, and agentic retrieval. agentmemory: Fix Memory Loss (30 sec)

Fix agent memory loss in 30 seconds with agentmemory. Prompt Engineering Techniques (2026) The prompting techniques that actually move the needle in 2026. MarkItDown: PDF to Markdown for RAG Convert PDFs, DOCX, HTML to clean Markdown for RAG pipelines. google/skills: Official Agent Skills Library Google's official Agent Skills for BigQuery, Cloud Run, GKE, Firebase. last30days: Real-Time Research Skill Real-time research across Reddit, X, YouTube, HN. Video Codebase Memory MCP: A Map for Agents Turn your repo into a code graph - half the tokens per task. Agent Modes: Plan vs Default vs Auto Three trust levels and the gear-shifting workflow between them. Agent Sandbox: OS-Level Security Filesystem + network walls: Seatbelt, bwrap, gVisor, Firecracker. From Prompts to Harness: 3 Shifts Why AI engineering moved from phrasing to information to control. Harness: The 6 Components Context, tools, orchestration, state, evaluation, recovery. Video Pi Extensions: Rewrite the Harness In-process extensions vs out-of-process hooks - what each can actually reach, and where the real ceiling is. Video Harness in Practice: OpenAI + Anthropic What OpenAI and Anthropic changed for production agent systems. YC's QM: A Source-Code Read Twice as much of YC's open-source harness governs scope as drives the model - the memory notebook, the audience-filtered transcript, the egress proxy, and the limits its own SECURITY.md names. Video Loop Engineering: Write Verifiers, Not Prompts Why the verifier is the bottleneck, open vs closed loops, and how to make either ship. Loop Engineering: The Anthropic Playbook Anthropic's gather-act-verify loop, and the five principles that make it reliable. Loop Engineering, Karpathy-Style How Karpathy's generation-verification loop and autonomy slider became loop engineering. Karpathy's LOOPS.md: What It Says and What's Verified What the circulating LOOPS.md file claims, why the rules are real loop engineering, and what we could verify about the attribution. The 4 Types of Agentic Loops Turn-based, goal-based, time-based, and proactive loops - which part of the cycle each hands off. Addy Osmani's Loop Engineering: The 5 Components Osmani's five loop components, the three debts, and why you still own the outer loop. Video The Evolve Loop: Loops That Rewrite Themselves How an evolve run rewrites a loop's own contract, state, and triggers - and the rule that stops it becoming busywork. Loop Engineering in Practice: 30 Days of Data A real two-loop content system with every number published - the run log, the quality-valve skips, and the failures. Video Graph Engineering From the Inside: Control Graphs NEW The three things "graph engineering" means, the two ways to enforce a control graph (code vs a skill), four patterns for cutting nodes, and two production graphs. Graph Engineering: Wire Loops into a Graph When one agent loop is not enough: nodes, edges, shared state, the frameworks, and when NOT to reach for a graph. Graph vs Loop Engineering: What Changed What genuinely changed versus what is a rebrand - a before/after build and a paradigm gut-check. Compound vs Loop Engineering NEW What separates Every's compound engineering (units of work that teach the next one) from loop engineering (operations that run themselves), and a 3-question sort for which jobs fit which frame. Graph or Loop: How to Decide The 5 signals you have outgrown a loop, the cost of a graph you did not need, and a 30-second decision tree. Is It Just LangGraph? Frameworks Compared LangGraph vs AutoGen GraphFlow vs Google ADK vs A2A - the pattern versus the implementations. The 5 Layers of AI Engineering Prompt, context, harness, loop, graph - the whole stack, and which layer is your bottleneck right now. Graph Engineering with Claude Code Subagents as nodes, the orchestrator as edges, the Agent SDK - wire an agent graph with the tool you already have open. Peter Steinberger's "Loops or Graphs?" Tweet Trace the first four days of the July 2026 wave while separating Steinberger's amplification from the phrase's earlier history. Andrew Ng's Agentic Patterns, Mapped to Graphs What Ng published, what the viral PDF posts claim, and where AI Builder Club's graph mapping begins. Graph Engineering and the Karpathy Loop What Karpathy's autoresearch loop actually does, why the linked 11-page PDF is real but not an Anthropic publication, and which of the six steps are worth climbing. Google Graph Engineering: Claim vs Reality Why the viral free Google course does not exist, the ADK 2.0 graph workflows that do, and the free Google material worth your time. How to Evaluate AI Agents Generator-evaluator, traces, eval sets, and LLM-as-judge pitfalls. Video Proactive Agents & the Company-as-Code Pattern What a 147-agent "AI company" repo teaches about proactive, self-improving systems. Claude Fable 5: Access, Pricing, Prompting How to use Claude Fable 5 - access, prompting, pricing, refusals. How to Become an AI-Native Company Which loop to build first, what earns autonomy next, what breaks. AI Agent ROI: The Fleet Audit The landed-not-produced audit: how to tell a working fleet from activity theater, with our own failure receipts. Reviewing AI-Generated Pull Requests The review packet, the AI_POLICY.md, and the three machine gates that run before a human sees the diff. Why Your Agent Bill Is Wrong The cost categories session metrics leave out, a script that refuses to print a partial total, and the file that stops recursive fan-out. A Role Label Is Not a Sandbox A two-phase canary suite that proves a deny rule actually denies, why denying one command leaves the other routes open, and how to read a tool grant for what is missing. Your Agents Have Production Credentials and No Owner A fillable agent registry, an offboarding runbook whose third step makes you watch the old credential fail, an append-only action-log contract with correlation IDs, and the autonomy ladder as a table with demotion triggers. How to Build an SEO Agent Loop The four loop roles search actually needs, the emerging-term wedge with our own Search Console numbers, and the rank-monitoring rule that scores the daily series, reads the incomplete-data boundary the API response carries, and refuses to produce a verdict from a trailing average. How to Cold-Launch a Social Presence With Agent Loops Seven social loops on one fleet, four of which reach a live account without a human in the moment, what each has measurably earned, the three ways a posting loop goes quiet while reporting ok, and the account-level cadence rule that takes the slot atomically instead of reading a ledger that cannot be read in time. AI Coding Tools Cursor, IDEs & the rest of the stack 1 Cursor & Tools 1 Cursor & Tools Video Ship Code 10x Faster with Cursor Complete Cursor IDE setup - Tab, Cmd+K, Chat, Composer, .cursorrules. Video The 5 Levels of Coding Agent Autonomy From simple code review to fully autonomous engineering. Google Antigravity: Complete Guide Complete guide to Google's free agent IDE. Video Codex CLI: How It Works + When to Use It The auth/billing fork, the 2026 model decision tree, and Codex vs Claude Code. Comparisons, role guides, tool reviews, and builder stories. Agent Skills Library Claude Code for Your Role Builder Stories & Guides Join AI Builder Club Get the free newsletter Weekly deep-dives on AI tools, automation workflows, and builder strategies. Join 5,000+ readers. No spam. Unsubscribe anytime.