Colorhive: 60fps 3D Hex-Stacking & Zero-GC Cascades in Unity
How we engineered Colorhive, a 3D hexagonal color-sorting puzzle game for Android: axial coordinate math, zero-allocation cascades and 60fps thermal budgets.
Read the piece ↗ Applied engineering
experience and perspectives
AI systems, delivery decisions, product operations and the mistakes worth writing down. Written by the senior team between engagements, not by a content calendar.
Mobile & Game Showcases
Colorhive: 60fps 3D Hex-Stacking & Zero-GC Cascades in Unity
How we engineered Colorhive, a 3D hexagonal color-sorting puzzle game for Android: axial coordinate math, zero-allocation cascades and 60fps thermal budgets.
Mobile & Game Showcases
Basic Todo App: Zero-Latency Offline-First Mobile Architecture
Why most task managers fail the 300ms thumb test, and how we engineered an instant, zero-telemetry offline-first React Native app with local atomic storage.
Orzed How It Works
Move faster with senior product and engineering guidance
How Orzed pairs senior product and engineering leads with AI accelerated execution. Where the senior call sits, where AI runs, how the Console surfaces both.
Orzed Models & Agents
Orzed QA Agent, evidence based validation
Technical write up of the Orzed QA Agent. Pulse based composite with deterministic test runner plus security and static analysis. Evidence backed verdicts.
AI Engineering
Why AI Integration Fails at the Governance Layer
The model is rarely the failure. A missing prompt registry, evaluation gate and human approval path are where production AI integrations actually fall over.
Agent Systems
Multi-Agent Handoff Patterns That Actually Work
Three handoff patterns for production multi-agent systems with the failure modes and observability hooks that decide whether the agents earn their keep.
Console Changelog
Automated QA Agents are now live
The first automated layer of the Orzed Review pipeline is live. Built on Orzed Pulse with public security and static analysis models, evidence backed verdicts.
Agent Systems
Agent Memory Design Beyond the Chat History
Three memory shapes for production agents: working memory, episodic memory, semantic memory. Picking the right one for the failure you are trying to fix.
LLM Cost Engineering
LLM Cost Routing: The Cheapest Passing Model Pattern
Routing each request to the cheapest model that passes its eval cuts inference spend 30 to 50 percent. Architecture, routing logic and eval discipline.
Orzed Models & Agents
Introducing Orzed Horizon, the flagship planning model
Orzed Horizon is the flagship model in the Orzed stack. Long context, deep reasoning, sized for planning, architecture and senior review work. Async by design.
Orzed Models & Agents
Introducing Orzed Meridian, the workhorse production model
Orzed Meridian is the middle tier of the Orzed stack. Sized for everyday coding, content and structured output, with strong tool use and function calling.
Orzed Models & Agents
Introducing Orzed Pulse, the always on lightweight model
Orzed Pulse is the lightweight tier of the Orzed stack. Distilled from Horizon, sized for QA gates, classification, routing and sub second validation.