Software architecture · AI · technical strategy

What Building Real Software With AI Actually Taught Me

Ten things a year of shipping AI-written production code actually taught me, minus the hype.

Use AI where probabilistic behavior creates leverage. Use deterministic systems where certainty is the feature.

Read my writing Work with me

Selected thinking

Ideas worth arguing about.

What Building Real Software With AI Actually Taught Me

Ten things a year of shipping AI-written production code actually taught me, minus the hype.

hiring2026-09-12

The Budgeting Mistake That's Quietly Wrecking Your Software Team

Companies price developer hires against a market salary band. What they should be pricing is expected impact within a fixed window — and the difference is where hiring budgets bleed out.

Why Navier-Stokes Actually Matters

OpenAI says an AI system proved a fluid equation can break down in finite time. Here's the actual Navier-Stokes problem, precisely.

AI Agents That Actually Work: How to Combine Judgment, Code, and Control

A practical case for combining probabilistic models with deterministic workflows and human judgment.

leadership2026-08-26

The Conflict That Moves a Team Forward: How to Disagree Without Breaking It

Why productive disagreement can reduce rework, improve decisions, and protect team trust without romanticizing conflict.

software architecture2026-08-19

Why Building Software Is Hard: The Cost of Ambiguity

Software fails in the gaps between goals, communication, system states, and real-world behavior. A founder-focused lesson from a real-time hardware integration.

Most AI Agents Should Not Be Agents

A practical boundary between useful probabilistic systems and workflows that need certainty.

engineering2026-08-19

Fast Shipping Is an Engineering Constraint, Not an Excuse

Shipping quickly is a technical requirement—but speed still needs judgment about irreversible risk.

performance2026-08-19

Optimize the Work Before Optimizing the Code

The highest-leverage performance optimization is often deleting work the system never needed to do.

architecture2026-08-19

Your Startup Probably Does Not Need Microservices

Microservices are a trade: exchange local complexity for distributed complexity only when the benefits are real.

What I work on

Problems that do not fit neatly inside one job title.

01

Architecture & technical strategy

System design, modernization, scaling decisions, and architecture rescue.

02

AI & automation

Workflow analysis, AI system design, retrieval, agents, and deterministic automation.

03

Performance & reliability

Measurement-led investigations into latency, cost, bottlenecks, and failure paths.

04

Architecture & technical strategy

System design, modernization, scaling decisions, and architecture rescue.

Principles

Engineering is mostly choosing which tradeoffs are acceptable.

The right answer depends on the problem, its constraints, and what happens when the system is wrong.

01

Ship fast, but understand the risk.

02

Simplicity is a feature.

03

Measure before optimizing.

04

AI is leverage, not authority.

05

Failure paths matter.

Currently studying

Agent architectureRetrieval systemsDistributed systemsPerformance engineeringSecurityAlgorithmsLLM evaluationAutomation economics

About

I am more interested in understanding the system than defending a tool.

I build software, study systems, and spend a lot of time asking why we build things the way we do.

I am Arnold Moya. I work across architecture, AI, performance, automation, and the operational details that decide whether software survives contact with reality.

Shipping speed, quality, performance, security, simplicity, and business value are not independent virtues. Good engineering is the tradeoff that fits the problem.

Consulting

Have a difficult technical problem?

I take on selected advisory and technical engagements.