Deep-dive library: for people building with Claude Code

A glossary of 300+ terms, the full history of AI, how agents differ from classic automation, and which AI devices are worth it.

1. ★
   AI Mayak Academy Glossary: 2026A glossary of 300+ terms about AI, Claude Code, programming and business, with plain-language pictures and links to lessons.
   User
2. ★
   The evolution of AI: 76 years from Turing to the agents of 2026A detailed 76-year history of AI: the eras, the winters, the key people and the road to agents.
   User
3. ★
   Agentic workflows vs. traditional automationHow agentic workflows differ from Zapier and n8n: where they win, and where classic automation is simpler.
   Confident user
4. ★
   The AI tool market's big stories: Devin, Windsurf and ManusThe stories of Devin, Windsurf and Manus from 2024 to 2026, and how to build a process that doesn't depend on one tool.
   User
5. ★
   AI Hardware in 2026: Apple Intelligence, Smart Glasses and AI Wearables — What to Buy and WhyWhich AI devices to buy: phones, smart glasses, wearables, laptops, and what they mean for privacy.
   Confident user

Install Claude Code, write a standing instruction for the agent, choose a work mode and learn the built-in commands.

1. ★
   The Four C's framework: how to structure your AI OSThe Four C's: context, connections, capabilities and cadence as the frame for a personal AI system on Claude Code.
   Confident user
2. ★
   Installation and setup: VS Code, Claude Code, first launchInstall VS Code and Claude Code, sign in, pick a plan and have your first conversation with the agent.
   Builder
3. ★
   CLAUDE.md: your project's system promptCLAUDE.md: a standing instruction file for your agent, three layers (WHAT, WHY, HOW) and how it saves tokens.
   Builder
4. ★
   Claude Code permission modes and CLI launch modesClaude Code launch modes and permission modes: Manual, Accept edits, Plan, Auto and Bypass.
   Builder
5. ★
   Claude Code built-in commandsClaude Code commands (/clear, /compact, /context, /init) and permission modes, from Plan to Bypass.
   Confident user

Project structure, safe storage for keys, a first live workflow, and what to do when it breaks.

1. ★
   The WAT framework: workflows, agent and toolsThe WAT framework (workflows, agent, tools): how to structure your project folders and improve the system over time.
   Builder
2. ★
   Setting up API keys and .env: a safe startHow to store API keys safely: .env files, .gitignore, wrangler secret and a password manager.
   Confident user
3. ★
   Build your first AI workflow live: newsletter automationYour first workflow from scratch: a news newsletter built with Perplexity, Claude and Gmail, with a human check before sending.
   Builder
4. ★
   Debugging and Self-RepairWhat to do when a workflow breaks: describe the symptom to the agent and let it find and fix the cause.
   Confident user

Tokens and the context window, why quality drops in long sessions, folders, knowledge bases and answering from your own documents.

1. ★
   Tokens and context management in Claude CodeWhat tokens and the context window are, how to read /context and how to save tokens with progressive loading.
   Confident user
2. ★
   Context Rot and 28 techniques to fight degradationContext rot: why quality drops as the window fills, and 28 techniques to prevent it.
   Confident user
3. ★
   Context management: advanced techniquesContext management: a short CLAUDE.md, subagents, /clear, /compact and loading files gradually.
   Builder
4. ★
   The philosophy of folders: the PARA system for an AI entrepreneurPARA and file-naming rules: how folders and CLAUDE.md make the agent's work more accurate.
   Confident user
5. ★
   Obsidian as a second brain for the AI eraObsidian as a knowledge base for Claude: folder structure, Hot Cache, plugins and connecting your vault.
   Confident user
6. ★
   Notion AI as your team's knowledge baseNotion as shared memory for a team and its agents: connecting the Notion MCP and three starter templates.
   Confident user
7. ★
   RAG: Retrieval Augmented GenerationRAG: how to teach Claude to answer from your own documents using embeddings and a vector database.
   Builder

Connect outside services, count what MCP really costs, build your own MCP server, work with images and PDFs, and deploy so it runs around the clock.

1. ★
   APIs and integrationsWhat an API is: requests, keys, response codes and first integrations with Stripe, Twilio and SendGrid.
   Builder
2. ★
   MCPs: expanding what Claude Code can doMCP: how to connect outside services to Claude Code, which servers you need and what they cost in context.
   Builder
3. ★
   MCP Pricing & Token Economics: free vs paid servers, the real cost of integrationsWhat an MCP stack really costs: a free server, the paid service behind it and the tokens it uses in context.
   Builder
4. ★
   MCP Builder: building your own MCP serverBuild your own MCP server in TypeScript or Python: connect it to Claude Code and test it in the Inspector.
   Engineer
5. ★
   Multimodal Claude (working with more than text): images, PDFs, audioImages and PDFs in Claude: how to share an image, what it costs in tokens and what the limits are.
   Builder
6. ★
   Deploying 24/7: how to put your project online with Cloudflare Workers, GitHub and VercelDeploying automations to Cloudflare Workers: cron, secrets, KV, limits and how it compares with other platforms.
   Builder
7. ★
   n8n + AI: smart workflows with LLM nodes and agentsn8n with AI nodes: self-hosting, the AI Agent, lead and content workflows, and license terms when you sell to clients.
   Builder

What skills are, how to build and test one, when to use ready-made skills, and a set of reusable prompt modes.

1. ★
   What are Skills (reusable instructions for Claude): expertise you can reuseWhat Claude Code skills are: a folder with a SKILL.md file, progressive loading and six steps to create one.
   Builder
2. ★
   Building a skill (a reusable instruction for Claude) from scratch, LIVE: Skill Creator + the Eval (automatic quality check) FrameworkBuild a skill with Skill Creator: its questions, testing it, tuning the description and iterating.
   Builder
3. ★
   Skill architecture (a skill is a reusable instruction for Claude): 2 archetypesTwo kinds of skills: ones that extend what Claude can do and ones that encode how you work, plus size and storage.
   Builder
4. ★
   Evals: skills that improve themselvesEvals for skills: tests, pass rate, checking that the skill triggers and a data-driven improvement loop.
   Engineer
5. ★
   Use, don't build: when to use a ready-made Claude skill and when to create your ownReady-made Claude Code skills and plugins: when to use someone else's, and when to write your own.
   Builder
6. ★
   Arsenal of Prompts: 6 reusable modesSix ready-made prompt modes (writer, editor, debug, research, architect, ADR) and how to build your own.
   Builder

Rules that fire on their own, and helper agents with their own context and model.

1. ★
   Hooks: automatic rules for Claude CodeClaude Code hooks: events, handlers, matchers and settings.json. Rules that fire on their own.
   Builder
2. ★
   Hooks LIVE: building hooks from scratchThree working hooks: protection from leaking keys, a daily limit on operations and a log of the agent's actions.
   Builder
3. ★
   Subagents: specialized workersSubagents: a separate context, tools and model for each task, with files in .claude/agents.
   Builder
4. ★
   Agent teams: working in parallelAgent teams: parallel work with a shared task list. An experimental feature, off by default.
   Engineer

Let the agent drive a browser or apps on screen, run tasks on a schedule, from scripts and in CI, and in parallel branches.

1. ★
   Browser automation: Playwright + QABrowser automation: Playwright and Claude in Chrome for testing forms and checking your own websites.
   Builder
2. ★
   Computer Use: controlling the screenComputer use: Claude operates apps on your screen. Access levels and safety rules.
   Builder
3. ★
   Loop vs Scheduled Tasks: when to use whichWhen to repeat a task inside a session with /loop, and when to use a schedule in the cloud, in the app or an outside cron job.
   Builder
4. ★
   Claude Code CLI: when the terminal beats the IDEClaude Code in the terminal: -p mode, pipes, scripts, worktrees and running it in GitHub Actions.
   Builder
5. ★
   Headless Mode and CI/CD: Claude without a UIClaude Code with no interface: the -p flag, --bare, JSON output, GitHub Actions and cost control in CI.
   Engineer
6. ★
   Git worktrees: parallel developmentGit worktrees and claude --worktree: parallel sessions on different branches without conflicts.
   Builder

Permissions, extended thinking, cheaper bulk requests, voice input, and community tips.

1. ★
   Permissions and securityClaude Code permission modes, secrets in .env and how to store keys in production.
   Builder
2. ★
   Extended Thinking: thinking things throughClaude's thinking mode in the API: adaptive thinking, the effort parameter and showing the reasoning.
   Engineer
3. ★
   Prompt Caching and the Batch API: economies of scaleHow prompt caching and the batch API cut the cost of high-volume requests to Claude.
   Engineer
4. ★
   Voice tools: Aqua Voice, /voice mode, WhisperVoice instead of keyboard: Claude Code's built-in /voice, dictation apps, Whisper and two-way voice.
   Confident user
5. ★
   32 Claude Code hacks: from beginner to power user32 tips for Claude Code at three levels: context, subagents, hooks, worktrees, permissions and scheduling.
   Builder
6. ★
   Ultra hacks 2026: finds from 45+ sources across the internetTen tips from the community: two Claudes, a retrospective, output compression, hooks and an honest take on how useful AI really is.
   Builder

Design tools for a product, voiceover and voice cloning, video editing with the agent, and a social media pipeline.

1. ★
   Design atlas: AI image generation, icons, animation and UI systemsDesign tools for a product: image generation, Figma MCP, UI libraries, icons, animation and video.
   Builder
2. ★
   AI voiceover: ElevenLabs, the TTS stack, voice contentTurning text into speech: ElevenLabs, voice cloning, OpenAI TTS and local models, with a pipeline from script to video.
   Builder
3. ★
   Video editing with Claude Code: FFmpeg, DaVinci Resolve MCP, RemotionEditing video with Claude Code: FFmpeg commands, a YouTube pipeline, DaVinci Resolve MCP, Remotion and batch processing.
   Builder
4. ★
   Social media automation: Instagram, Telegram, content on autopilotA content pipeline for Instagram and other social channels: the official API, schedulers, text generation and comment replies.
   Builder

A personal assistant for your day, bots for Slack, Teams, WhatsApp and Telegram, intent routing, and Claude inside Microsoft tools.

1. ★
   AI Executive Assistant: a capstone assistant projectA personal AI assistant: a daily summary, meeting prep and an email review that run on a schedule.
   Builder
2. ★
   AI in messaging apps: Telegram, Slack, WhatsApp and Teams on autopilotClaude chatbots for Slack, WhatsApp, Teams and other messengers: routing, handoff to a human and a no-code option.
   Builder
3. ★
   Telegram bots with the Claude API, from zero to productionA chatbot with the Claude API from zero to production: setup, polling and webhook, memory, voice and deployment on Cloudflare Workers.
   Builder
4. ★
   Chatbot managers: NLP, intent routing and smart handoff to a humanA chatbot with intent routing: Claude Haiku, rules, handoff to an operator and metrics.
   Builder
5. ★
   Microsoft 365 + Claude: the full integrationClaude in the Microsoft world: Teams webhooks and bots, Word, Excel, PowerPoint, Graph API, Power Automate and Foundry.
   Builder

Ticket answers from a knowledge base, a chat widget, phone support agents and real-time voice.

1. ★
   AI Customer Support: a ticket system, RAG over a knowledge base, smart supportCustomer support with Claude: ticket classification, answers from a knowledge base, escalation to a human and metrics.
   Builder
2. ★
   AI chat on your website: Crisp + Claude, a smart widgetA Claude chat widget on your website: a Crisp webhook, handoff to a human, RAG over your docs and question analytics.
   Builder
3. ★
   Call Support AI: Vapi + Bland.ai, voice calls in customer supportVoice agents for support calls: Vapi, Bland.ai, a knowledge base, call review and an honest cost-per-minute calculation.
   Builder
4. ★
   Voice AI agents: Vapi, Retell, Bland.ai, real phone callsVoice agents on Vapi: the call, a knowledge base, a webhook with Claude analysis, cost per minute and the legal rules for calls.
   Builder
5. ★
   Real-time AI: streaming, WebRTC and low-latency conversational agentsStreaming Claude answers: SSE, WebSocket, TTFT metrics, LiveKit and OpenAI Realtime for voice.
   Engineer

Lead pipelines, product funnels, a CRM that updates itself, search and AI-answer visibility, product analytics and competitor monitoring.

1. ★
   Finding potential clients: from scraping websites to a CSV fileA lead pipeline: Apify, filtering, website analysis with Claude and an export to CSV.
   Builder
2. ★
   Funnel design for AI products: from visitor to paying customerAn AI product funnel: five stages, three archetypes, where it leaks and a worked example for learning.
   Builder
3. ★
   CRM on autopilot: HubSpot + AI, deals without a sales repHubSpot and Claude: an email becomes an updated customer record, a lead score from 0 to 100 and a follow-up task.
   Builder
4. ★
   The SEO machine: the full stack with Claude CodeA stack for traffic from search and AI answers: keywords, briefs, schema markup and a weekly audit in Claude Code.
   Builder
5. ★
   Product analytics with AI: PostHog, Mixpanel and smart insights without a data analystProduct analytics with AI: PostHog and HogQL, an automated weekly report, churn prediction and A/B tests.
   Builder
6. ★
   AI competitive intelligence: automatic market monitoring with ClaudeCompetitor monitoring: Playwright, ChangeDetection and Claude collect prices and pages into a weekly report.
   Builder

Autonomous agents, sandboxes for running code safely, browser agents, multi-agent frameworks and advanced computer use.

1. ★
   Manus AI: autonomous agents for tasks that take hoursHow the autonomous agent Manus differs from Claude Code, when to pick each and how to give it tasks.
   Confident user
2. ★
   AI sandboxes: E2B and Modal for running agent code safelyRunning agent code safely in a sandbox: E2B and Modal, analyzing a CSV, a comparison and the cost.
   Engineer
3. ★
   Browser agents: Browserbase, Stagehand and web automationBrowser agents with Browserbase and Stagehand: act, extract, observe, price monitoring and the ethics of scraping.
   Engineer
4. ★
   Multi-agent orchestration: LangGraph, CrewAI and Mastra vs. Claude CodeMulti-agent systems: LangGraph, CrewAI, Mastra and Claude Code subagents, and when you need a framework.
   Engineer
5. ★
   Advanced Computer Use: patterns for complex automationComputer use for complex scenarios: the screenshot loop, retries after failures, isolation and a hybrid with Playwright.
   Engineer

Open models on your machine, when fine-tuning makes sense, local agents, and an internal AI platform for a team.

1. ★
   Local AI models: Ollama, LM Studio and private AIOpen models on your own computer: Ollama, LM Studio, choosing by memory, an LLM router and a document processor.
   Builder
2. ★
   Fine-tuning: when prompts aren't enoughWhen a prompt is enough, when you need RAG and when you need fine-tuning: a decision tree, LoRA and where to train.
   Builder
3. ★
   Local AI agents: minimal agents, OpenClaw, Hermes, Qwen-AgentLocal agents without the cloud: Ollama, OpenClaw, Hermes, Qwen-Agent and smolagents. When local beats an API.
   Builder
4. ★
   Self-hosted AI Enterprise Stack: vLLM, SGLang, Docker. An internal AI platform for your teamAn internal AI platform for a team: vLLM, SGLang, LiteLLM, Docker, SSO and monitoring. When self-hosting pays off.
   Engineer

A catalog of product ideas and payment models, ways to cut model costs, and unit economics at three business sizes.

1. ★
   Product and Business Atlas: MCP Connector Catalog, Professions and Product IdeasA catalog of MCP connectors, an atlas of professions and product ideas with payment models, with no income forecasts.
   Builder
2. ★
   Cost engineering: $20 vs. $200 a month, or how to pay 10 times less for the same resultsNine ways to cut LLM costs: model choice, caching, batch, local models, limits and monitoring.
   Builder
3. ★
   Unit economics of an AI stack: the full financial math at 3 levels (solo / SMB / enterprise)AI stack unit economics: LTV, CAC, margin and payback on example numbers for a solo builder, a studio and a team.
   Confident user

Why protection matters before you launch, taking payments, keeping secrets safe, AI regulation, and logs of what agents did.

1. ★
   Why a business owner needs security and architectureFour loss scenarios that happen without security and architecture, and a route through the seven protection lessons.
   Builder
2. ★
   Payments & Billing: Stripe MCP, subscriptions, webhooksTaking payments and subscriptions: Stripe MCP, payment links, webhooks, recovering overdue payments and MRR metrics.
   Builder
3. ★
   Security in Claude Code: .env, secrets and data protectionHow to prevent secret leaks: .env, .gitignore, Claude Code deny rules, a pre-commit hook and secrets in production.
   Confident user
4. ★
   AI regulation & compliance 2026: the EU AI Act, GDPR, Anthropic's terms, and what small businesses should watch forAI regulation as of October 2026: the EU AI Act, GDPR, US state laws and a small-business checklist. A general overview, not legal advice.
   Builder
5. ★
   Managing an army of agents: logs, oversight, ClickUp and CRMAgent logs: JSONL, STATUS.md, Supabase, ClickUp and Notion through MCP, a CRM and an audit trail in git.
   Builder

Tracing and prompt tests, what to monitor in production, recovering from typical failures, backups, and the full architecture.

1. ★
   MLOps for indie builders: monitoring, drift and retraining without a DevOps teamMLOps without a DevOps team: call tracing, prompt regression tests, cost control and prompt versioning.
   Engineer
2. ★
   Production observability: what to monitor when your agent is liveWhat to monitor on a production agent: five metrics, three alert levels, tools, debugging from a complaint and a dashboard.
   Engineer
3. ★
   Failure recovery patterns: what to do when your agent breaks in productionSeven common ways AI agents break in production, how to recover, chaos tests and incident reviews.
   Engineer
4. ★
   Backups and recovery for your AI stack: what to do when the Anthropic API is downBackups and disaster recovery for your AI stack: backup providers, the 3-2-1 rule, drills and message templates.
   Engineer
5. ★
   The final architecture: a business's complete AI stackThe architecture of an AI business as a whole: five stack layers, an automation layer, Microsoft integration and cost by stage.
   Builder

Checking plugins before and after install, protecting hooks from the agent itself, and defending against prompt injection.

1. ★
   Plugin Security: 9 attack categories and a trust registryPlugin security: nine categories of attack, checks before and after installing, and a registry of trusted plugins.
   Engineer
2. ★
   Hook-Deny-By-Design: hooks that protect themselvesHow to stop the agent from changing its own Claude Code hooks: four layers of protection and a channel for permission.
   Engineer
3. ★
   Prompt injection defense, the top security threat of 2026Prompt injection: six attack types, eight layers of defense, red team tests and an incident response plan.
   Engineer

A knowledge base the agent can navigate, learning paths over shared knowledge, project templates at three sizes, and projects you can detach and sell.

1. ★
   Knowledge Atlas: a map of knowledge in 54 layersA knowledge atlas: a network of short layers with an INDEX.json, so the agent finds what it needs in the fewest tokens.
   Engineer
2. ★
   Curricula: 8 learning tracks for different readersHow to build learning around a shared knowledge base: eight tracks for different readers instead of a single course.
   Builder
3. ★
   3-tier templates: solo, mid, corporateThree tiers of project template (solo, mid-size, corporate) and the modules you add only when you really need them.
   Builder
4. ★
   Portfolio detachability: projects you can unplugHow to build projects you can detach and sell without breaking the platform: four red lines.
   Engineer

Algorithmic trading on a paper account and where AI meets crypto, with scam warning signs. Neither lesson is investment advice.

1. ★
   Algorithmic trading with Claude Code: connecting Alpaca through MCP and managing riskAlgorithmic trading on a paper account: Alpaca MCP, risk rules and a simple strategy. Not investment advice.
   Builder
2. ★
   Crypto + AI in 2026: what works, what's overhyped, and how not to lose moneyFive places where AI and crypto meet: what has substance, where the scams are and how to vet a project. Not advice.
   Builder

Week-long builds: a chatbot, a SaaS MVP, a content pipeline and a voice agent for a small business. The fifth project, AI consulting, is in the core course.

1. ★
   Build-Along: a Telegram bot for a business in 7 daysA week building a business chatbot on Cloudflare Workers with Claude, saved conversation state and Stripe billing.
   Builder
2. ★
   Build-Along: a SaaS MVP from idea to first usersA week building a SaaS MVP on Next.js, Supabase, Stripe and the Claude API, from customer interviews to launch.
   Builder
3. ★
   Build-Along: Content Factory, a content pipeline in 7 daysAn AI content pipeline: brand voice, prompts, images and voiceover, a publishing schedule and approvals.
   Builder
4. ★
   Build-Along: a voice agent for a business in 7 daysA voice AI agent for a small business with Vapi, Twilio and Claude: from choosing a niche to launching for a client.
   Builder