Building an Agentic Coder
How to build an autonomous agentic coder that leverages advanced LLMs to transform software development. A talk given at the Ottawa Forward JS meet-up.

This is the personal blog of Ben Houston (bio here), a 3D computer graphics software developer. For a quick overview of the work I am proud of, see my portfolio. I've worked on projects like Web3DSurvey, Three.js, Deadline, Krakatoa, and Exocortex VFX plugins. I'm also an entrepreneur, having founded Land of Assets, Threekit, Clara.io, Exocortex, and Frantic Films Software.
How to build an autonomous agentic coder that leverages advanced LLMs to transform software development. A talk given at the Ottawa Forward JS meet-up.
As AI dramatically reduces software development costs, we need to rethink our approach to code. Software is becoming ephemeral - easily replaced rather than maintained. This paradigm shift means optimizing for current needs instead of hypothetical futures.
AI agents are moving beyond IDEs into full automation of coding workflows. Discover how integrating agentic coding into GitHub Actions can streamline software development -- turning issues into PRs and automating documentation, tests, and even code reviews, all through simple GitHub comments.
Adding GitHub Mode to MyCoder has radically improved my productivity, making my agentic coder feel like a true team member. By leveraging GitHub for task tracking, pull requests, and CI/CD feedback, I’ve seen a 3x to 5x boost in development efficiency.
Adapting your codebase to match AI coder assumptions doesn't just help AI -- it uncovers hidden complexity, ultimately benefiting human developers.
As AI coding agents become increasingly capable, we need to rethink our code organization and documentation practices. Here's why traditional "human-first" coding approaches are becoming obsolete and how to structure your projects for optimal AI compatibility.
Why I have been using TanStack/Start instead of Remix or Next.js for recent React projects.
Open Core combines the viral growth of open source with the revenue potential of enterprise software – but most founders get the balance wrong. Drawing from my experience building both open source and enterprise products, I explain why this model particularly shines for developer tools, and how to execute it without alienating either developers or paying customers.
How I wrote the first version of Krakatoa at Frantic Films to render Doc Bailey's SPORE images for Stay, and how the renderer later reached Avatar, Harry Potter, and Exocortex Fury.
In 2002, I joined Frantic Films to write fluid simulation software. A small scheduler I built for those sims became Deadline, a render farm manager used by Blizzard, major VFX studios, and considered by the Academy for Sci Tech Awards.
After learning the hard way that chasing startups and SMBs was a dead end, I discovered why the path to sustainable SaaS growth lies in mastering enterprise sales and the 'land and expand' strategy.
How a pioneering web-based 3D editor pivoted from visual effects to enterprise product visualization.