I build applications with AI and shape the development process through custom skills, agents and disciplined use of Git.
My toolkit includes ChatGPT/Codex, Xcode, Zed, LM Studio, Antigravity, Hermes, T3 Code and PaperClip. I combine these tools with reusable instructions, change review and verification, retaining responsibility for the application.
Explore my GitHubI create my own skills to capture the instructions, context and repeatable steps I want AI tools to follow. They let me shape the development process around the work, rather than start every task from a blank prompt.
The aim is to make my approach reusable: clear expectations, relevant context and a consistent way to work through recurring tasks.
I use Hermes as part of my AI development workflow, alongside my editors and other AI tools.
It forms part of the way I work with agents and bring AI assistance into my development process.
I make extensive use of Git to track changes and manage the evolution of my applications. Version history provides a clear record of what changed and a way to revisit earlier decisions.
Working with AI tools and agents makes that discipline particularly valuable. I use Git to inspect changes, keep work reviewable and retain control over what becomes part of the application. My GitHub profile provides access to my public repositories.
I use agents to support development tasks, with instructions and context that define the work. Custom skills help make those expectations explicit.
I remain responsible for the application: reviewing changes, checking behaviour and deciding what is ready to use. Agent output is part of the development process, with review and verification still required.
OpenAI’s conversational AI and coding-agent tools support the path from exploring a requirement to implementing, testing and reviewing software changes.
In my workflow, custom skills and clear instructions give that work structure. I set the direction, review the changes and assess the evidence before deciding what is ready to use.
Apple’s development environment for building, testing and debugging applications across its platforms. Xcode is part of my toolkit for native application development.
Its role in my workflow is to bring implementation and validation together: inspecting the application, checking its behaviour and reviewing build and test results before deciding it is ready to progress.
Apple’s Xcode overviewAn open-source development environment that combines direct work on code with optional AI assistance.
For me, Zed keeps hands-on development and review close together. It provides the working environment for inspecting changes, refining implementation and retaining direct control over the application.
My on-premises AI engine for running local AI models, providing a local environment for experimentation and application development.
In my workflow, it brings model evaluation closer to the application: exploring suitability for a task, reviewing outputs and considering the operational requirements of locally hosted AI. I retain responsibility for the data used, access decisions and validation of results.
Google’s agentic development platform, designed for AI agents to plan, execute and verify software development tasks. It brings agent coordination and reviewable outputs into the development process.
In my toolkit, it supports a shift from individual coding prompts towards directing work, reviewing evidence and retaining accountability for the result.
Google’s platform overviewAn open-source interface for directing coding agents, bringing agent-driven development into a shared control surface across desktop, web and mobile.
Its relevance to my workflow is oversight: directing development tasks and reviewing the work of coding agents while retaining responsibility for what ships.
T3 Code project overviewAn open-source orchestration platform for teams of AI agents, bringing tasks, roles, permissions and work history into a coordinated operating structure.
From an executive perspective, its value is making agent work accountable: defining responsibilities, maintaining visibility and connecting individual tasks to a shared objective.
PaperClip project overviewMy public repositories are on GitHub. Each project’s own documentation is the best place to understand its purpose, setup and current state.
github.com/andypduk