The Mental Stack: What Street Fighter Can Teach Us About AI-Driven Development

Your team's newest developer has mass-committed 47 files across 12 libraries, three state managers, and a date parsing package you've never heard of. Oh — and that developer is an AI agent.

Every framework, library, and abstraction your team adopts adds to the "mental stack" — a concept I've stolen from fighting games. Street Fighter II launched with eight characters and a handful of special moves. Street Fighter 6 has drive impacts, perfect parries, burnout states, and a roster of 29. Both are similar, but the cognitive load is wildly different. Software works the same way.

The mental stack isn't just a human problem anymore. Just as a Street Fighter player can only track so many meters at once, your AI agents are 'reasoning' through a finite context window. Every unnecessary abstraction is a layer they can hallucinate through. A lean mental stack—**minimalism as a technical requirement**—means your agents write better code and your developers move faster.

We'll put this to the test with a decision every frontend team faces: server-side or client-side? We'll trace a single feature through each approach — i18n, forms, accessibility, state management — and stack up the true cognitive cost. I'll share case studies of when this went right and when it went catastrophically wrong.

You'll walk away with a framework for evaluating complexity trade-offs — for your team, for the developers who come after you, and for the agents now writing code alongside you. And if nothing else, you might learn a bit about Street Fighter too.

Taming the Agentic Storm: Securing OpenClaw with "The Claw Machine"

It’s your worst nightmare: your CTO just discovered OpenClaw on social media and wants it deployed by Monday. While the potential for productivity is massive, the security reality is a "curl | bash" nightmare that could bring your entire organization crumbling down. From rank-and-file engineers to C-Level executives, everyone wants a piece of the agentic future—but no one wants the breach that comes with it.

I found myself in this exact spot in February and realized we didn't just need a policy; we needed a platform. Enter The Claw Machine: an open-source project designed to wrap, sandbox, and govern autonomous agents.

This talk introduces The Claw Machine and demonstrates how to use it to bridge the gap between "Shadow AI" and enterprise-grade security. We will dive into how to leverage its core primitives to define strict network isolation, automated filesystem state management, and the deployment of "ClawBots"—secure, isolated instances of OpenClaw and its derivatives.

In this session, you will learn:
- The OpenClaw Threat Vector: Why standard security fails when agents start "thinking" for themselves.
- The Claw Machine Architecture: How the project uses sandboxing to create a "blast radius" for every agent.
- Implementing ClawBots: A step-by-step guide to configuring network egress rules and filesystem backup/restore cycles within the tool.
- Why it works: The infrastructure primitives that make the Claw Machine and its ClawBots secure.

By the end of this talk, you’ll be ready to deploy The Claw Machine to your organization and deliver the power of agents—safely. Just don't let your CFO see the token bill.

This talk is an overview of how to use my open source project, The Claw Machine. You can find more about it at https://theclawmachine.dev

It Finally Works on My Machine: Scaling Local-First CI for the Age of AI

Platform and DevOps Engineers, let’s face it: your CI pipelines weren’t built for the age of AI. As developers pair with autonomous agents, the volume of code is skyrocketing. These agents are prolific, but they are also messy—pushing commits that ignore linting rules, drift from templates, and incinerate your cloud build budget.

To survive this "agentic" era, we need to shift the center of gravity back to the workstation. This session introduces Local-First CI: a strategy to give developers (and their AI counterparts) immediate feedback without the "push-and-pray" loop.

In this talk, we will explore:
- The architecture of a robust local pipeline that mirrors your cloud environment.
- Strategies for an incremental rollout that doesn't disrupt existing workflows.
- The "Trust Gap": How to transform local checks into a legitimate gate rather than a suggestion.
- Live Build: We’ll build a Local-First CLI interface to show you exactly how to deliver your organization from CI purgatory.

GitHub Actions - Zero to Hero

A Comprehensive Guide to Modern CI/CD

GitHub Actions has evolved significantly since its introduction, developing a rich feature set that rivals established tools like Circle CI and Jenkins. After six years building CI systems across multiple platforms, I can confidently say that GitHub Actions stands out as the superior tool for modern development workflows.

This comprehensive session takes you from complete beginner to advanced practitioner, covering everything you need to know to leverage GitHub Actions effectively.

What We'll Cover

Foundation: Understanding the GitHub Actions Ecosystem
- Core concepts: workflows, jobs, steps, and triggers
- The strategic advantages of GitHub Actions over alternatives
- Setting up your environment for success
- Secrets management and security best practices

Operational Excellence: Day Two Considerations
- Cost management: understanding and optimizing GitHub Actions pricing
- Self-hosted runners: when, why, and how to implement them
- Scaling strategies for growing organizations
- Resource management for efficient pipeline execution

Getting Started: Building Your First Workflows
- Creating a beginner-friendly workflow from scratch
- YAML best practices for maintainability
- Workflow validation techniques
- Effective logging and notification strategies

Advancing Your Skills: Beyond Basic YAML
- Transitioning workflows to Golang or TypeScript for improved testability
- Implementing unit tests for your CI/CD code
- Developing reusable, composable actions
- Building custom actions for specialized requirements

Organizational Strategy: Standardization and Sharing
- Creating and distributing workflow templates
- Building an action snippet library for your organization
- Implementing governance and standards
- Designing effective role-based access controls

Expert Techniques: Complex Workflow Architecture
- Designing and implementing dynamic matrices
- Building sophisticated multi-stage pipelines
- Managing complex dependencies efficiently
- Deployment strategies for various environments

Takeaways

Whether you're just starting with GitHub Actions or you've been using them since the beta release, you'll leave with practical techniques to elevate your CI/CD capabilities. You'll understand how to build pipelines that are both reliable and flexible, meeting the needs of modern development teams.

By the end of the session, you'll have a comprehensive understanding of GitHub Actions from basic concepts to advanced implementations, with strategies you can immediately apply to improve your development workflows.

Come prepared to learn - we're covering a lot of ground!