Deterministic AI development patterns derived from the MetaCurtis Engine (15k+ particles @ 60 FPS).
This repo documents the core architectural and workflow patterns I use to make AI-generated code predictable, safe, and production-grade.
These patterns are the backbone of my AI-Native Pilot offering.
- Pattern S — Single Writer Governance
- Pattern D — Deterministic Directives
- Pattern C — Contracts & Evidence
- Pattern E — Eventguard Ordering
Each pattern has its own doc under patterns/.
Under protocol/ you’ll find the Kodex–Claude Protocol v3.0:
- Multi-agent AI workflow
- Evidence-driven debugging
- 10–15× feature velocity
- Deterministic, repeatable development patterns
Tools like Copilot, Cursor, and Claude make it easy to generate code…
…but teams are now dealing with:
- race conditions
- regressions
- architectural drift
- nondeterministic behavior
- unstable releases
These patterns and protocols are how I stabilized a complex WebGL engine — and how I stabilize AI-heavy systems in client codebases.
Engine repo: https://github.com/metacurtis/metacurtis-v3.1
Live demo: https://metacurtis.com
I work with teams that are:
- Using AI tools heavily in development
- Shipping complex frontends, WebGL, real-time UIs, or dashboards
- Struggling with instability from AI-generated code
In a 4-week AI-Native Pilot, I:
- Fix one painful engineering problem
- Apply deterministic patterns from this repo
- Leave the team with docs + workflows they can reuse
👉 Details: https://curtiswhorton.com
👉 CTO overview: FOR-CTOS.md