AI coding agents don't just suggest code anymore. They run shell commands, edit files, install packages and call the network on your behalf.
AI coding agents don't just suggest code anymore. They run shell commands, edit files, install packages and call the network on your behalf.
On your laptop, a single bad tool call or prompt injection is enough to delete your files, read your SSH keys and API tokens, or send your source code anywhere.
In this hands-on one-hour workshop, you'll see why "YOLO mode" on your own machine is a real risk, and how Docker Sandboxes (sbx) contains it. You'll learn how sbx runs an agent in an isolated sandbox that only sees your project folder, how its network policy controls which domains the agent can reach, and how its proxy injects credentials, so your API keys never actually enter the sandbox.
We'll put it into practice with Docker Agent and a Mistral AI model. You'll define an agent in YAML, start it in a sandbox, store the Mistral key securely, and watch what the agent can and can't do. You'll leave knowing how to give an agent full autonomy without giving it your machine.
You start a task in Claude Code, hit the usage cap halfway through, and switch to Codex. The plan, the edits, the tests, and the two approaches you already ruled out all stay behind. The new agent starts cold and walks into the same dead end. Most of us end up carrying that context between agents by hand.
In this hands-on session, you'll work on a real repo with Omnigent, an open source meta-harness that lets you run Claude Code, Codex, and other coding agents from one UI and CLI, so the context travels with the task instead of with you. You'll:
- set up a workflow where a task moves from one agent to another without losing its context
- add spend caps and approval policies, so you're not approving every action yourself
- write a custom agent as a YAML spec
- open a shared session that a teammate can review or take over from their own laptop
By the end of the hour, you'll have Omnigent running on your laptop with your own coding agents connected.
What to bring: a laptop with at least one coding agent you can sign in to, such as Claude Code, Codex, Cursor, or OpenCode, using either a subscription or an API key.
Please let me know if it works. I can change some parts if needed.
Are you interested in deploying a GitHub repository using Infrastructure as Code (Terraform) and learning how to set up the corresponding cloud-based development environment with GitHub Codespaces? Then join my workshop!
What you'll need:
A laptop. Ideally, bring your personal laptop to avoid any potential policy restrictions imposed by your company.
A personal GitHub account. You must be able to create a Personal Access Token (PAT), and you will be performing experiments on your GitHub account using Infrastructure as Code (IaC).
A current version of Visual Studio Code.
Please ensure the following extensions are installed:
Remote Explorer
GitHub Codespaces
Git installed and configured correctly. For example, you should be able to clone newly created repositories from your GitHub account using your correct identity information (name and email address).
AI can already write code, analyze data, create content, conduct research, and build prototypes in minutes.
As execution becomes faster and more accessible to almost anyone, an important question emerges: what makes our work valuable?
In this talk, we go beyond prompts and explore the skills and tools that become increasingly important as AI becomes capable of doing more.
We’ll look at critical thinking, creativity, expertise, and human judgment and how AI can amplify these abilities.
Because when AI can produce more, faster, the difference lies with the person who knows what is worth building, what is worth keeping, and which decisions are worth making.
Have you ever wondered what it takes to move GenAI from simple chat interfaces to production-grade systems in banking? Are you curious about how agentic AI and multi-agent architectures work under the hood in a highly regulated financial environment?
This masterclass explores generative AI in banking through two real-world use cases. In one, we look at how specialized agents collaborate, using frameworks like LangGraph, to support document-heavy Source of Wealth assessments. In the other, we look at a conversational retail-banking assistant, showing how a routing agent coordinates specialized domain agents and tools to handle everyday banking requests safely and compliantly.
Join us and let’s build together!
The cloud bill is up. Again.
Your service is slow and gets randomly killed by the OOM.
Before you start the great Rust rewrite, there's a more practical approach.
In this hands-on workshop, you'll learn how to make your Go code run faster and consume less memory.
More importantly, you'll learn a tried and true optimization process.
What you'll walk away with:
- Know *when* to optimize, and more importantly when *not* to optimize
- Create benchmarks that don't lie to you
- Take out the guesswork of finding the slow parts using a profiler
- How to make your code faster, from micro optimization up to new design
You'll work on a real Go service and make it faster. To follow along you need to have:
- The Go SDK
- Git
- Curiosity