AI-assisted Engineering Workshop Series

What you'll learn

This AI-Assisted Engineering training delivers practical guidance on using coding agents - and other AI-powered tools - to increase your productivity across a variety of tasks - not just in coding.

We'll review the core mechanics of LLMs, identifying some key limitations and how to work around them. We'll then apply that knowledge to build up a working model of your new AI assistant's "personality traits" - how its encyclopedic public knowledge contrasts with its inability to learn on the job, for example.

Throughout the series, we'll come back to AI's strengths and weaknesses as we identify where these tools can add the most value to your daily work. We'll also dive deep on the limitations of coding agents specifically, and learn a set of techniques which work around these limitations to get the most out of your AI assistant.

The format is a three-part series of live online workshops, delivered weekly over Zoom. The sessions build on each other, with a week in-between to digest the content and try out what you're learning.

Attendance is intentionally limited to a small group of experienced engineers. Expect diagrams, whiteboarding and lots of demos, with plenty of time for discussion and Q&A.

Session 1 (60 minutes)

LLM Foundations

We'll review the fundamentals of how a Large Language Model works, and build an intuition around it's central limitations: hallucinations, and an inability to learn or form memories.

We'll then see how systems like chatbots and agents work around those limitations using context engineering.

Topics include:

• Token completion - the core of the LLM
• From tokens to chatbots
• From prompt engineering to context engineering
• Why hallucinations happen
• RAG (retrieval augmented generation)
• Tools and agents

Session 2 (90 minutes)

Real-world AI-assisted Engineering

We'll take a tour through various ways that AI can help with the day-to-day of building software. Yes, that includes help with coding, but our examples will extend to broader use-cases.

We'll spend most of our time in live demos and real-world examples, for use cases like design and prototyping, building internal tools, ad-hoc tasks in the terminal, and more.

We'll also discuss sweet-spots and danger zones for AI in these different use cases, building on what we learned in week 1 about the limitations of LLMs.

Use cases include:

• Writing code
• Work chores
• Understanding code
• Prototyping & Design
• Internal tooling
• Reviewing code
• Technical Research

Tools we'll use include:

• Copilot
• Claude Code
• Gemini 3 Pro / Nano Banana Pro
• v0
• ChatGPT Deep Research

Session 3 (90 minutes)

Agentic Coding

Coding agents (Claude Code, Codex, Copilot Agent, ...) are the key to real productivity gains with AI. Our final workshop will focus on practical guidance on getting your coding agent to do what you actually want.

I will lay out two central principles for succeeding with agents - context management and better feedback loops - along with specific techniques which implement those principles.

Topics include:

• Anatomy of an agent
• Succeeding with agents: constraining the solution space
• Context management
• Better feedback loops
• Incremental planning

Praise from past attendees

"I really appreciated that the content was presented in a way that didn't make me feel like I was being sold something. It felt really refreshing to hear it presented as "It's here, it's cool and useful, but there are barbs that you need to work through"

"Awesome workshop!"

Meet your instructor

I'm Pete Hodgson, a 25 year veteran of the software industry. I've spent many years consulting with engineering teams, helping them adopt new technical practices.

Despite my initial skepticism around AI, when coding agents came onto the scene I saw clear utility in my own daily development work. AI's not a magic wand, but it certainly makes me a more productive engineer.

In contrast, I've been frustrated that the overwhelming narrative around AI for software development has been one of extremes - either all software will be AI-generated in the next 2 years, or AI is total snake oil.

For the last couple of years I have attempted to walk a middle ground, offering practical, hype-free guidance on how AI can act as a "mech suit" for engineers - topics like why your coding agent is getting it wrong (and how to fix it), managing coding agent context, and building incrementally with coding agents.