# Trust Agent > Free AI course downloads. Workbooks by Mickarle Wagstaff-Irons - Micky Irons, published by Mickai LTD. Mickai is a British company building a Sovereign Intelligence Operating System. Each published course in the current free library is a PDF workbook with a matching web page, a worked Q&A and a review date. No account is needed. ## About this site - Publisher: Mickai LTD (Companies House 17166618), https://mickai.co.uk - Byline: Workbooks by Mickarle Wagstaff-Irons - Micky Irons, published by Mickai LTD. Workbooks use AI assistance; each course page shows its sources and review date. - [About](https://trust-agent.ai/about.html): why the workbooks are free, how they are made and the people behind them - [Library](https://trust-agent.ai/library.html): every workbook by level and goal - [Start here](https://trust-agent.ai/start.html): a short quiz that suggests a learning path - [Learning paths](https://trust-agent.ai/learning-paths.html): practical routes, prerequisites and limited compliance coverage; twenty premium courses and referral credits are planned, not open for purchase or use - [Free labs](https://trust-agent.ai/labs.html): a sandboxed HTML, CSS and JavaScript playground, a tokeniser lab and a sampling lab that run in the browser - [Certificate](https://trust-agent.ai/certificate.html): a self-declared certificate of completion for any workbook, made on the visitor's device - Sitemap: https://trust-agent.ai/sitemap.xml; feed: https://trust-agent.ai/feed.xml ## The Mickai network - Mickai (https://mickai.co.uk): the publisher. A British company building a Sovereign Intelligence Operating System. - Unified (https://unified.vip/): a journal on sovereign AI, audit and regulation. - Common Fortune (https://community-fortune.vip/): the economics of digital and AI sovereignty, with sourced figures. ## Courses - [What is AI? A plain-English starter](https://trust-agent.ai/courses/what-is-ai.html): A jargon-free introduction to modern AI: how chatbots work, what they are good and bad at, and how to use them safely from day one. (level 1, about 75 minutes, reviewed 2026-09-25) - [Your first prompts: how to ask AI for what you actually want](https://trust-agent.ai/courses/your-first-prompts.html): Learn to ask AI for what you actually want: the six parts of a good prompt, how to improve a reply, and how to build a personal prompt library. Free tools only. (level 1, about 120 minutes, reviewed 2026-09-25) - [The free AI toolkit: how to choose tools and stay safe](https://trust-agent.ai/courses/free-ai-tools.html): How free AI tools work, how to choose and test them, how to check their current terms and privacy, how to spot fake apps and scams, and how to build your own starter stack. (level 1, about 120 minutes, reviewed 2026-09-25) - [Make YouTube videos with AI: a beginner's workflow](https://trust-agent.ai/courses/ai-youtube-videos.html): A beginner's end-to-end workflow for making YouTube videos with AI: script, voiceover, visuals, captions, upload, YouTube's disclosure rules, analytics and a 30-day plan. (level 1, about 140 minutes, reviewed 2026-09-25) - [Stay safe with AI: scams, fake media and your privacy](https://trust-agent.ai/courses/ai-safety-everyday.html): Spot AI-assisted scams and fake media, keep your privacy when using chatbots, know your basic UK data rights and where to report problems, with a simple routine to follow. (level 1, about 100 minutes, reviewed 2026-09-26) - [Use AI to study, research and write without losing your own thinking](https://trust-agent.ai/courses/ai-for-study-and-writing.html): Use an AI assistant as a study partner, not an answer machine: quiz yourself, verify every claim and citation, write in your own voice, and say how you used it. (level 1, about 110 minutes, reviewed 2026-09-26) - [Using AI to find a job: your CV, cover letter and interview prep](https://trust-agent.ai/courses/ai-for-job-search.html): A practical, free workbook for using AI in a job search: tightening a CV against a specific job description, drafting and personalising a cover letter, and rehearsing interview answers, with honest limits on what AI can and cannot verify for you. (level 1, about 70 minutes, reviewed 2026-09-29) - [Using AI for everyday admin: letters, emails and forms](https://trust-agent.ai/courses/ai-for-everyday-admin.html): A practical, free workbook for the AI use case almost everyone actually has: writing a clear email or letter for something that matters, and filling in a form correctly. Covers tone, structure, and the checks you still have to do yourself. (level 1, about 55 minutes, reviewed 2026-09-29) - [Using AI as an accessibility tool: reading, writing and seeing](https://trust-agent.ai/courses/ai-for-accessibility.html): A practical, free workbook on the AI-powered accessibility features most computers and phones already have: screen readers that read a page aloud, dictation that types what you say, live captions for video and calls, and image description for photos a screen reader cannot otherwise interpret. Covers what each does well, where it gets things wrong, and how to check its work. (level 1, about 55 minutes, reviewed 2026-09-29) - [Build your first app with an AI coding assistant](https://trust-agent.ai/courses/first-app-with-ai-coding-assistant.html): Build a small bill-splitting web page with any AI coding assistant while staying in control: brief it, build in steps, review the code, test edge cases and keep safe copies. (level 2, about 120 minutes, reviewed 2026-09-26) - [Git for AI builders: commits, branches and pull requests](https://trust-agent.ai/courses/git-collaborative-development.html): Learn Git from your first commit to a reviewed pull request, with safe habits for AI-written code: read every diff, commit small, undo safely and keep secrets out. (level 2, about 125 minutes, reviewed 2026-09-26) - [HTML: build a meaningful, accessible page](https://trust-agent.ai/courses/html-accessible-page.html): Build a real event page by hand with headings, landmarks, alt text, a data table and a labelled form. Check it with the W3C validator and a keyboard, then review AI-written HTML. (level 2, about 130 minutes, reviewed 2026-09-26) - [AI images and video: rights, provenance, labels and quality checks](https://trust-agent.ai/courses/image-video-workflows-rights.html): Use AI images and video responsibly: ownership in the UK, EU and US, tool terms, consent, labelling, Content Credentials, quality checks and simple records. (level 2, about 105 minutes, reviewed 2026-09-26) - [Running open-weight AI models on your own computer, for free](https://trust-agent.ai/courses/run-open-models-free.html): A practical, hands-on guide to running open-weight AI models on your own computer at no subscription cost, using Ollama, LM Studio or llama.cpp, with honest guidance on hardware limits and quantisation. (level 2, about 65 minutes, reviewed 2026-09-29) - [Which AI model should I use? Picking the right AI for the job](https://trust-agent.ai/courses/which-ai-model.html): How to choose between AI assistants and models: what to weigh, how to run a fair five-prompt bake-off, and how to read leaderboards and benchmarks without being fooled. (level 2, about 120 minutes, reviewed 2026-09-25) - [Build and launch a website with AI](https://trust-agent.ai/courses/build-a-website-with-ai.html): Plan a simple site, have AI write the HTML and CSS, understand and check the code, then publish it free with HTTPS and sound search, accessibility and privacy basics. (level 2, about 120 minutes, reviewed 2026-09-25) - [Automate a spreadsheet workflow with AI, checks and an audit trail](https://trust-agent.ai/courses/spreadsheet-automation-audit.html): Clean a messy spreadsheet with AI-suggested formulas without losing control: work on a copy, test every formula, log each change and reconcile the totals. (level 2, about 120 minutes, reviewed 2026-09-26) - [What is an AI agent? Agentic AI explained](https://trust-agent.ai/courses/ai-agents-explained.html): A plain-English guide to AI agents: how they differ from chatbots and workflows, the parts and the loop, levels of autonomy, the main risks and how to keep a human in control. (level 2, about 75 minutes, reviewed 2026-09-25) - [CSS: responsive layouts, grids and typography](https://trust-agent.ai/courses/css-responsive-layout.html): Style a page by hand so it reads well on any screen: box model, cascade, flexbox, grid, fluid type, container queries, dark mode and focus, all measured at three widths. (level 2, about 140 minutes, reviewed 2026-09-26) - [AGI explained: what it means and what it does not](https://trust-agent.ai/courses/agi-explained.html): A balanced guide to artificial general intelligence: the main definitions, one published levels framework, how to read headlines and timelines, and what matters in practice. (level 2, about 75 minutes, reviewed 2026-09-25) - [Python: from first script to a useful automation](https://trust-agent.ai/courses/python-first-automation.html): A complete beginner's route from installing Python to a tested, dry-run-first tool that sorts and reports on a folder, using only what comes with Python. (level 2, about 140 minutes, reviewed 2026-09-26) - [Code for as little as possible: VS Code, OpenCode and app builders](https://trust-agent.ai/courses/code-cheaply-vscode-opencode.html): Work out what AI-assisted coding really costs and keep it low: how tools charge, a token cost calculator, free editors, OpenCode, local models and paid app builders. (level 2, about 125 minutes, reviewed 2026-09-27) - [The loop harness: how AI agents actually run](https://trust-agent.ai/courses/loop-harness.html): How an agent harness runs a model in a loop: context, tools, permissions, budgets, stop conditions, logging and testing, with a minimal Python loop and a paper design exercise. (level 3, about 90 minutes, reviewed 2026-09-25) - [Build your first chatbot](https://trust-agent.ai/courses/build-a-chatbot.html): Build a chatbot that behaves: model access, system prompt, history and context limits, guardrails, tone, a test set, logging and cost, on a no-code path or a code path. (level 3, about 105 minutes, reviewed 2026-09-25) - [Document pipelines: parsing, cleaning and chunking](https://trust-agent.ai/courses/document-pipelines-parsing-chunking.html): Build a local document pipeline: clean UTF-8 text, create traceable chunks, test boundaries and measure which evidence survives together. (level 3, about 120 minutes, reviewed 2026-09-27) - [Design a repeatable evaluation suite for an AI application](https://trust-agent.ai/courses/evaluation-suite-ai-application.html): Build an offline AI evaluation harness with test cases, a rubric, regression comparisons, uncertainty checks and a reproducible release record. (level 3, about 95 minutes, reviewed 2026-09-27) - [Tokenisation: train and inspect a small tokenizer](https://trust-agent.ai/courses/tokenisation-tiny-tokenizer.html): Train a byte-pair encoding tokeniser in plain Python, prove the round trip is lossless, and see how vocabulary size, languages, numbers and spaces change token counts. (level 3, about 105 minutes, reviewed 2026-09-27) - [Embeddings and vector search from first principles](https://trust-agent.ai/courses/embeddings-vector-search.html): Build vector search from scratch: turn text into vectors, compare them three ways, search by brute force, score it with recall@k and MRR, and see where it fails. (level 3, about 95 minutes, reviewed 2026-09-26) - [Host a chatbot on your website, safely](https://trust-agent.ai/courses/host-a-chatbot.html): Put a chatbot on your site without leaking keys or running up bills: server function, limits, CORS, privacy, accessibility, fallbacks, monitoring and a launch checklist. (level 3, about 90 minutes, reviewed 2026-09-25) - [Prompt injection: threat analysis and defensive testing](https://trust-agent.ai/courses/prompt-injection-defensive-testing.html): Learn why prompt injection happens, analyse the harm it can do, and measure which defences work with a harmless offline Python lab that uses fake canary secrets. (level 3, about 135 minutes, reviewed 2026-09-26) - [SQL: model and query your first database](https://trust-agent.ai/courses/sql-first-database.html): A beginner's route through SQL with SQLite: tables, keys and constraints, queries, joins, safe changes, indexes, injection-safe Python and reviewing AI-written SQL. (level 3, about 140 minutes, reviewed 2026-09-26) - [Test-driven development with an AI coding assistant](https://trust-agent.ai/courses/tdd-with-ai-coding-assistant.html): Use failing examples to guide an AI coding assistant, test boundary cases, catch deliberate defects and hand over a small Python change with clear evidence. (level 3, about 150 minutes, reviewed 2026-09-27) - [Tensors, gradients and a first training loop](https://trust-agent.ai/courses/tensors-gradients-training-loop.html): Learn what training is by doing it on a CPU: tensors and shapes, the loss, gradients by hand and by autograd, the training loop, validation and the classic bugs. (level 3, about 95 minutes, reviewed 2026-09-26) - [Data analysis with Python: investigate and explain a dataset](https://trust-agent.ai/courses/data-analysis-python.html): Answer a real question from a messy table with pandas: inspect, clean with a log, group, join, chart, measure uncertainty and check AI-written analysis code. (level 3, about 90 minutes, reviewed 2026-09-26) - [GPU memory maths for LLMs and SLMs](https://trust-agent.ai/courses/gpu-memory-maths.html): A repeatable method for checking whether a language model fits your GPU: weights, KV cache, training memory and adapters, plus a small calculator you can run. (level 3, about 75 minutes, reviewed 2026-09-25) - [JavaScript: the browser, events and application state](https://trust-agent.ai/courses/javascript-browser-events-state.html): Add behaviour to the event page with plain JavaScript: the DOM, events and delegation, one state object, form validation, focus, status messages and localStorage. (level 3, about 140 minutes, reviewed 2026-09-26) - [Node.js: build and test a small API](https://trust-agent.ai/courses/nodejs-small-api.html): Build a local read-only HTTP API with Node.js: explicit routes, strict query validation, predictable errors and executable tests using built-in modules. (level 3, about 150 minutes, reviewed 2026-09-27) - [Privacy and data minimisation in AI applications](https://trust-agent.ai/courses/privacy-data-minimisation.html): Keep personal data under control in an AI feature: UK GDPR principles, DPIAs, roles and transfers, and a redaction pipeline you run and test to see what it misses. (level 3, about 100 minutes, reviewed 2026-09-26) - [RAG and knowledge bases, step by step](https://trust-agent.ai/courses/rag-knowledge-bases.html): Build a knowledge base a model can answer from: collect, clean, chunk, embed, retrieve, cite and evaluate, with security basics and a hands-on exercise using five short documents. (level 3, about 100 minutes, reviewed 2026-09-25) - [Tool calling and structured outputs: validate every response](https://trust-agent.ai/courses/tool-calling-structured-outputs.html): Treat every model reply as untrusted input: write JSON Schema contracts, parse and validate each tool call, retry with feedback, and run tools with limits, dry runs and logs. (level 3, about 95 minutes, reviewed 2026-09-26) - [Go: build a concurrent service](https://trust-agent.ai/courses/go-concurrent-service.html): Build a standard-library Go HTTP service with bounded workers, ordered results, input limits, cancellation and executable tests. (level 3, about 180 minutes, reviewed 2026-09-27) - [Open model licences, datasets and deployment obligations](https://trust-agent.ai/courses/open-model-licences-datasets.html): Decide whether you may use, host and redistribute an 'open' model or dataset: licence families, dataset rights, model cards, AI-BOMs, hash checks and a checklist tool. (level 3, about 110 minutes, reviewed 2026-09-26) - [TypeScript: types for safer web applications](https://trust-agent.ai/courses/typescript-safer-web-apps.html): Build a typed reading-list boundary with runtime validation, explicit display states, compiler checks and executable tests. (level 3, about 180 minutes, reviewed 2026-09-27) - [AI governance: risk registers, evidence and human accountability](https://trust-agent.ai/courses/ai-governance-risk-registers.html): Govern an AI system without being an engineer: map hazards, score and evidence them, name owners and stop rights, make oversight real, and check the register with tested code. (level 3, about 115 minutes, reviewed 2026-09-26) - [Testing agent loops: budgets, termination and failure recovery](https://trust-agent.ai/courses/testing-agent-loops.html): Test an agent loop locally: enforce budgets, replay model decisions, reject late results and verify bounded retries without paying for model calls. (level 3, about 150 minutes, reviewed 2026-09-27) - [Inference engines explained](https://trust-agent.ai/courses/inference-engines.html): How inference engines run a model: memory maths, quantisation, prefill and decode, batching, hardware, and serving a model safely over a local API. (level 4, about 80 minutes, reviewed 2026-09-25) - [Reasoning models and reasoning engines](https://trust-agent.ai/courses/reasoning-models.html): What reasoning models are, when extra thinking pays off, how to control it, why a reasoning trace is not proof, and how classical reasoning engines differ. (level 4, about 80 minutes, reviewed 2026-09-25) - [How to evaluate a new frontier model the day it lands](https://trust-agent.ai/courses/evaluate-frontier-models.html): A ten-step method for judging a new model: primary sources, licence, architecture, hardware, benchmarks, your own tests, safety, jurisdiction and a decision record. (level 4, about 150 minutes, reviewed 2026-09-25) - [Hybrid search and reranking for better retrieval](https://trust-agent.ai/courses/hybrid-search-reranking.html): Implement reciprocal rank fusion, validate a reranking boundary and compare query-level retrieval results using a small reproducible Python lab. (level 4, about 180 minutes, reviewed 2026-09-27) - [Throughput, latency and batching: capacity planning for a model server](https://trust-agent.ai/courses/throughput-latency-batching.html): Build a tested batch queue simulator, compare burst and sparse arrivals, calculate timing metrics and design a measured model-server experiment. (level 4, about 180 minutes, reviewed 2026-09-27) - [Transformers: attention explained and implemented](https://trust-agent.ai/courses/transformers-attention.html): Implement scaled attention, test causal visibility and assemble a small forward-only transformer block using standard-library Python. (level 4, about 180 minutes, reviewed 2026-09-27) - [Evaluate a RAG system with a grounded test set](https://trust-agent.ai/courses/evaluate-rag-systems.html): Build a reproducible RAG evaluation fixture, score retrieval separately from claim support and completeness, and detect regressions hidden by a single headline number. (level 4, about 180 minutes, reviewed 2026-09-27) - [Build a tiny language model from scratch](https://trust-agent.ai/courses/tiny-language-model-from-scratch.html): Train an original character-bigram language model with explicit gradients, held-out evaluation, bounded sampling and a reusable JSON checkpoint using standard-library Python. (level 4, about 180 minutes, reviewed 2026-09-27) - [Model routing, caching and graceful fallbacks](https://trust-agent.ai/courses/model-routing-caching-fallbacks.html): Build a scoped response cache and bounded model router, test expiry and failure boundaries, and compare cost, latency and answer quality on a replayable trace. (level 4, about 180 minutes, reviewed 2026-09-27) - [Permission-aware retrieval and private knowledge bases](https://trust-agent.ai/courses/permission-aware-retrieval.html): Build and test a small retrieval policy that separates tenants, checks document groups before ranking and refuses missing permission matches. (level 4, about 180 minutes, reviewed 2026-09-27) - [Rust: ownership, safety and a small command-line tool](https://trust-agent.ai/courses/rust-first-cli-ownership.html): Build a standard-library Rust command that reads bounded fictional token counts, validates records and reports totals, with ownership examples and executable tests. (level 4, about 180 minutes, reviewed 2026-09-27) - [Harness state and agent memory: checkpoints, retention and deletion](https://trust-agent.ai/courses/harness-state-and-memory.html): Build a small SQLite state store with versioned checkpoints, inspectable facts, expiry, tenant-scoped deletion and tested recovery using standard-library Python. (level 4, about 180 minutes, reviewed 2026-09-27) - [Knowledge graphs and structured retrieval](https://trust-agent.ai/courses/knowledge-graphs-retrieval.html): Build a small directed knowledge graph in Python, retrieve bounded evidence paths and test direction, cycles, missing facts and source traceability. (level 4, about 180 minutes, reviewed 2026-09-27) - [Threat-model an agent and its tool boundary](https://trust-agent.ai/courses/threat-model-agent-tool-boundary.html): Map an agent tool boundary and test a local gateway for identity, ownership, exact-action approval, expiry, replay and stale writes using fictional notes. (level 4, about 180 minutes, reviewed 2026-09-27) - [Orchestrating multiple agents with explicit responsibilities](https://trust-agent.ai/courses/multi-agent-orchestration.html): Build a tested Python coordinator with scripted research, draft and review roles, checked handoffs, dependency failures, bounded retries and a shared call budget. (level 4, about 180 minutes, reviewed 2026-09-27) ## How to cite Cite a course by its title, Mickai LTD and the course URL. The PDF workbooks are free to download and share unchanged; please link to the course page rather than copying it.