AI for Makers and Solopreneurs

Learn how AI-assisted coding is a guided collaboration, not a vending machine, via a four-step loop—provide context, receive output, evaluate, and redirect—as you build in-browser RC filter analyzer with Claude.

Explore the five levels of AI-assisted coding and how the Shapiro framework guides oversight. Choose the right level, from reference to agent, for each task.

Frame prompts with four elements: task, context, constraints, and output format. This disciplined communication shows how precise specifications prevent gaps and bad results.

Discover how to build a complete prompt with four elements - task, context, constraints, and output format - using a frontier model interview to generate a project specification via SpecBuilder.

Identify failure modes in artificial intelligence like hallucination, context drift, overconfidence, and wrong calculations, then apply mitigations such as verifying imports and committing changes before tasks.

Create a self-contained RC filter analyzer in the browser using claude.ai, from a written specification. Visualize the cutoff frequency with a log-frequency magnitude plot using R and C inputs.

Set up and use Visual Studio Code as the default editor, navigate the explorer and terminal, and manage local versus cloud workflows with extensions like Klein and Claude Code.

Explore how the client acts as an autonomous coding agent with tools to read, write, run commands, and browse the web, driven by the agentic loop of plan and act.

Demonstrate cloud code and Claude code as coding assistants, contrasting terminal and GUI workflows, installation, VS Code usage, and practical planning and editing examples.

Learn how llms act as a separate engine from the interface, using context windows and three tiers: fast, capable, frontier, for cost-conscious, efficient work.

Compare local and cloud LLMs, weighing VRAM and hardware realities against cloud benefits like latency and updates. Assess privacy, offline use, and cost, explaining why the course favors cloud-first workflows.

Translate your brainstormed milestone into a precise specification defining task, context, constraints, and output format for coding agent. Initialize git before AI and run a handoff test to reduce ambiguity.

Follow a planning approach to scaffold a serial plotter project by forming a four-element specification, creating a git-based structure, configuring serial settings, and aligning dependencies with Flask and PySerial.

Practice targeted iteration from revision 0, diagnose issues with three-part follow-ups (symptom, component, desired outcome), and commit fixes as a build record in y-axis, NAN mean, pause resume burst.

Build the complete front-end for the serial plotter with index.html, style.css, and app.js, using html5 and chart.js from a CDN to render a live, multi-axis dashboard from the backend stream.

Document for future you by writing intentional docstrings and a candid readme that explains how to run, how the project is structured, and why key design choices exist.

Implement a logger class wired into app.py to persist Arduino data to timestamped csv files in a log directory, gated by browser connection, plus module docstrings and a readme.

Switch the application from mock data to real Arduino data via the serial port. Generate one hertz csv data (timestamp, temperature, humidity, lux) with an Arduino Uno R4 minima and upload sketch.

Explore when to deploy a runtime LLM versus a coding assistant, weighing unstructured input, generation needs, and cost, latency, and privacy for building a datasheet Q&A tool.

Explore planning a runtime llm powered datasheet qa tool, validate the architecture with Claude, apply a decision test, and outline a rag-based pipeline for pdf data sheets.

Learn how retrieval augmented generation (rag) handles large documents like datasheets by extracting relevant sections through chunking, embedding, and indexing to guide reliable answers.

Build the full retrieval-augmented generation pipeline by enabling PDF upload, text extraction, and document-based question answering so the model cites uploaded data.

This lecture completes the rag implementation by naive pdf text extraction, chunking with a 900–950 splitter, embedding via mini llm l6 v2, and storing in chroma db, with ingest.py integration.

Explore the retrieval step in a rag implementation by building an ask.py that embeds a question, retrieves the top eight Chroma chunks, and builds a prompt with those chunks.

Convert pdf to Markdown to preserve table structure for data sheets, improving RAG ingestion and retrieval accuracy by maintaining correct tables and values.

Examine cost, latency, and failure in AI projects for makers and solopreneurs. Analyze how cost, latency, and failure affect AI projects for makers and solopreneurs.

Explore resilience and cost considerations for AI projects, guiding makers and solopreneurs to build robust solutions while managing resources and trade-offs.

Build a large multifile booking system with an agent, emphasizing authentication, resource CRUD, conflict detection, and per-user booking history within a structured skeleton; apply memory.md and Cloud Code discipline.

Learn to depend on external APIs and implement background jobs with graceful handling. Use naga.date to check public holidays, cache results, and serve data when the API is unavailable.

Close the capstone by delivering functional, stable, documented, and deployed software; validate with testing and deployment to a real URL, and prepare a developer README and a member guide.

Learn how solopreneurs monetize with merchant-of-record payments, handle VAT and GST, set up checkout and transactional emails, and plan production deployment, landing pages, and user documentation.

Set up the database schema and scaffold a Flask app with SQLAlchemy models for user, resource, booking, and reminder sent, plus a plan for directory structure and memory tracking.

Implement administrators' resource management interface, including list, add, edit, and archive actions, with protected routes, unit tests, and data seeding, guided by prompts and migrations.

Enhance the booking system's user interface with widened tables, a modern look, header and footer improvements, and admin capabilities to edit, archive, and manage resources, including bookings.

Implement the audit log, the members page with edit, and the bookings admin pages; add resource thumbnails and admin-only upload; run tests and commit changes as you plan future features.

Develop a comprehensive developers readme, track versions with git and automatic bumps, and differentiate deployment runbook from readme.md while planning the VPS deployment.

Explore deploying an AI project to a VPS in this part 1 module for makers and solopreneurs, outlining setup basics and practical deployment steps.

Deploys a VPS-based web app, starts the Kadi service, secures the site with Let's Encrypt certificate, seeds data, and enables backups, Cloudflare protection, and backwards-compatible migrations with rollback workflow.

Explore practical multi-tenancy for makers and solopreneurs, detailing tenant-aware routing, password reset email strategies, and Cloudflare-based subdomain provisioning with migration planning.

The lecture walks through completing chunk 3 of a multi-tenancy feature, detailing plan, files created and modified, a migration node, verification tests added, and progress tracking with 255 tests passed.

This chunk refactors the multi-tenancy project with Flask blueprints, resolves root route conflicts by renaming routes for public and member, and prepares six new files with tests.

Overhaul makerbooking.com's front-end user interface, fix routes, and replace the audit log with a modern timeline view. Add a statistics page and enhanced resources across admin, tenant, and home pages.