Private workshop · 1 hour · live online
AI coding agents, explained for game programmers.
A one-hour walk-through for engineers who haven’t tried these tools yet. We cover what AI agents can do today, how Claude Code works under the hood (model selection, effort levels, ultracode, token use), and which parts of game programming they suit. Unreal Engine is front and center, though the ideas carry to any tool and any engine.
illustrative · what an agent session looks like
Why now
AI changed how software gets written. Game code is catching up.
The demos that go viral are built on a to-do app. They fall apart the moment you point them at a few million lines of engine code, long build times, Blueprints, and performance budgets that don’t forgive mistakes.
So this isn’t a hype reel. In one hour we give your team a clear-eyed view: what these agents are good at, how Claude Code works once you look past the marketing, and where it fits in real game development. The ideas stay tool-agnostic. The examples lean on Unreal.
What your team takes away
Walk out knowing where AI earns its place.
What AI agents can (and can’t) do
An honest look at where these tools help today, and where they still trip up.
How Claude Code actually works
A plain-English tour of the controls that matter: picking a model, setting effort levels, ultracode, and why token use changes what you get back.
Where it fits in game dev
Which tasks are a good fit (boilerplate, tooling, tests, refactors) and which to keep in human hands.
Unreal Engine, specifically
Real UE5 examples across C++, Blueprints, and the editor, so it’s clear how this works on an engine and not a toy project.
How to set it up well
Setup and configuration tips so anyone on the team can start without fighting the tool.
Guardrails that matter
IP and data handling, performance-critical code, and how to review AI output without rubber-stamping it.
Beyond writing code
The payoff isn’t always new code.
Writing features is the obvious use. Two of the most useful turn out to be quieter: making sense of code your team didn’t write, and keeping documentation that doesn’t go stale.
Get your bearings in unfamiliar code
Point the agent at a system nobody fully remembers (an inherited project, a third-party plugin, the subsystem whose author left years ago) and ask how it works. It can trace a call path, map what depends on what, or explain what a module does and why. New hires get up to speed in days. Veterans finally get answers about corners of the engine they’ve never opened.
Documentation that keeps up with the code
Once it understands a system, it can write it down. The agent drops Markdown straight into your repo: architecture notes, onboarding guides, decision records. Open those files in Obsidian and you get a linked, searchable map of the project, backlinks and all. It’s the documentation nobody has time to write, produced as a side effect of the work.
The hour
How the hour tends to go.
A live walk-through and demo, nothing to install. We keep it loose and follow the room. Examples lean on Unreal Engine 5, though the ideas apply to any tool and any engine.
Where things stand
What today’s AI coding agents can do, and what they can’t. A clear starting point, whatever tool you land on.
Inside Claude Code
A tour of the controls that matter: model selection, effort levels, ultracode, and how token use shapes the results.
Reading and documenting code
Using the agent to understand an unfamiliar codebase, then turn that understanding into Markdown docs and an Obsidian vault.
Where it fits in game dev
Which systems and tasks suit AI and which to keep human-owned, with concrete Unreal Engine examples.
Setting it up
Configuration that works, plus the guardrails: IP and data, performance-critical code, reviewing what the agent writes.
Questions
Open floor, shaped around your engine, your codebase, and how your team works.
Who it’s for
- Gameplay, tools, and engine programmers
- Tech leads and engineering managers
- Anyone who hasn’t tried AI coding tools yet
- Technical directors sizing up AI tooling
What to expect
- A live walk-through and demo, not a hands-on lab
- Nothing to install; your team just joins the call
- Plain English, no prior AI experience needed
- Real time set aside for your questions
Format & logistics
The shape of the hour.
Who’s running it
Studio-to-studio, not a vendor pitch.
Run by the team behind Game Dev Outsourcing. We work with game studios every day and use these tools on real game code ourselves, so this comes from practice, not a slide deck about the future. Expect straight answers: what works today, where it bites, and how to bring it into a team without the wheels coming off.
Questions
Before you ask.
Do we need AI experience to take part?+
None. The session is built for people who haven’t tried these tools, and it starts from the ground up in plain English.
Is it really just an hour? Can it go deeper?+
It’s a focused hour. If your team wants a longer, hands-on follow-up afterwards, we can set that up separately.
Will this apply to our engine and stack?+
The examples use Unreal Engine 5 and C++, but the ideas are engine- and tool-agnostic. Unity, Godot, and in-house engines all apply, and we tailor the examples to your stack.
Is our source code or IP exposed?+
This is a walk-through and demo, so we’re not touching your repo. We do cover how these tools handle code, what stays private, the data controls on offer, and safe habits, so your team can make its own call.
What do attendees need on the day?+
A calendar invite. It’s a live talk and demo, with nothing to install and no prep.
Can the session be tailored to us?+
Yes, and it should be. Tell us your engine, your sore spots, and what your team is curious about, and we build the hour around it.
How much does it cost?+
It depends on team size and format. Register your interest below and we’ll send options.
Bring this to your team
Register your team’s interest.
Tell us a bit about your studio and we’ll follow up to scope the hour and pick a date. No commitment. This just starts a conversation.
- One focused hour, shaped around your engine
- No setup; your team just joins the call
- Real time for your team’s questions