BLOG

Thoughts on Building

Writing about AI-augmented development, shipping SaaS as a solo developer, and the tools that make it possible.

All posts

Idempotency: The Skill That Quietly Separates Senior Engineers

A code-first masterclass on idempotency, the art of writing software that is safe to run twice. Prevent duplicate charges, corrupt data, and broken retries across frontend, backend, distributed systems, and AI agents.

Distributed SystemsAPI DesignBackend
18 min read

I Spent a Whole Commit Making Invisible Koi 'Brighter'

A koi shader named a local `patch` — a reserved GLSL word — so the fragment shader silently failed to compile and the fish rendered nothing. I spent a commit trying to make the 'faint' markings bolder. Why tsc and next build stayed green, and the console probe that finally caught it.

WebGLShadersDebugging
8 min read
FEATURED

Prompt Engineering Is Now a Subset of a Bigger Job

The craft moved from wording one message to curating the whole context window. Why reasoning models inverted the old prompting advice — 'think step by step' and few-shot can now hurt — and the disciplines that replaced clever wording.

AI DevelopmentContext EngineeringPrompt Engineering
14 min read
FEATURED

ChatGPT vs Claude vs Gemini vs Copilot vs Meta AI: A Developer's Honest 2026 Comparison

A primary-sourced, June-2026 comparison of ChatGPT, Claude, Gemini, GitHub Copilot, and Meta AI — real strengths, real weaknesses, models, context windows, pricing, and which to use for coding, writing, multimodal, and research. From a developer who ships with them.

AI ModelsAI ToolsLLM Comparison
17 min read
FEATURED

Your AI Agent Forgets and Degrades — Engineering Context Across Time

An AI coding agent is bounded by what it can hold in context — and that context rots within a session and evaporates between them. The research on context rot, and the file-based memory architecture that makes a months-long project get smarter instead of resetting.

AI DevelopmentContext EngineeringAI Agent Memory
13 min read
FEATURED

AI Writes 1.7× More Bugs. Two Disciplines Decide Whether They Ship.

AI-generated code carries ~1.7× more defects — and the worst ones pass every test. The two engineering disciplines that catch them before users do: verification gates that measure output, and agent orchestration that contains errors instead of amplifying them.

AI DevelopmentEngineering DisciplineCode Review
12 min read

I Shipped a Blog Post No One Could Read

A scroll-reveal animation left an entire article stuck at opacity: 0 — invisible to readers, crawlers, and no-JS clients. The IntersectionObserver threshold bug behind it, why it only struck the longest post, and the progressive-enhancement fix you can copy.

Progressive EnhancementSEOAccessibility
9 min read

Building a WCAG AA Contrast Budget into Your CI Gate

How I treat accessibility contrast as a build invariant — the same way you treat bundle size. Three real CI catches across two months, one bug that would have shipped to retina users invisibly, and the GitHub Action pattern you can copy.

AccessibilityCI/CDWebGL
11 min read

How I Ship SaaS Products 10x Faster with Claude Code

How one developer delivers what agencies charge 3x more for — and why that means faster timelines, lower cost, and a better product for the people who hire me.

AI DevelopmentClaude CodeProductivity
4 min read

Master Prompts: The Architecture Behind AI-Driven Development

The planning method behind every project I build — why it means zero scope creep, no surprises, and a final product that matches exactly what you asked for.

AI DevelopmentPrompt EngineeringArchitecture
4 min read

Building PowderLedger: From Rental Shop Floor to Production SaaS

How a real operational problem became a production SaaS that replaced spreadsheets, paper forms, and three separate tools — from concept to live customers.

Case StudySaaSPowderLedger
4 min read
OPEN TO WORK
Text (845) 544-8757
fastest response
hello@alejandro.software
replies within 24h
Book a call · 30 min →