Coding agents and local models
Applied AI
I use AI to write and check software, not only to chat. Some of it runs on my own machine.
How I build with agents
- Review the work: I ask for review and run automated checks before shipping changes.
- Write the playbook once: Repeatable instructions so the next run does not start from a blank chat.
- Use local models for experiments: Small models on this machine for drafts; other models when a task needs them.
What I am experimenting with
Long coding sessions
For long coding sessions, I keep credentials in a separate secrets workflow.
Interactive browser prototypes
Testing 3D, shader and motion ideas in the Lab before bringing the best interactions into client work.
Automated quality checks
Pairing agent-assisted code with automated layout, accessibility and motion checks before shipping.
The agents, models and platforms behind it are on the Stack page.
Code activity
19,792 commits to the Zyta codebase since Mar 2026 · updated 18 November 2026
- Busiest day
- 21 September 2026 · 809 commits
- Active days
- 246
- Avg. commits per active day
- 80.5
19792 commits to the Zyta codebase over the last 37 weeks; busiest day was 2026-09-21 with 809 commits.