Blog

Building in Public

Real stories from the trenches — how we use AI automation to solve actual business problems, what results we get, and what we learn along the way.

Featured February 26, 2026 · 12 min read

How a 3-Person Team Handles 200+ Leads Per Month with AI

A travel agency was drowning in WhatsApp messages and Excel spreadsheets. We built a complete digital storefront with ML lead scoring, AI chat, automated workflows, and voice agents — all on Cloudflare at $0/month hosting. Now they handle 5x more leads without hiring.

GM

Gonzalo Monzón

Latest Articles

Fresh insights from the trenches

Engineering
10 min read

Vanilla JS Is the Assembly Language of the Browser (And That's Why We Use It)

Gmail ships 20MB of JavaScript. Slack ships 55MB. Our cookie consent system? 4KB, zero dependencies. When you understand the low-level primitives — vanilla JS, pure CSS, raw DOM APIs — you don't need a framework to tell you what the browser already knows. But we're not framework haters: we use Astro for SSG and React islands where they genuinely help. The difference is that we choose our tools — they don't choose us.

GM
March 13, 2026
Read Article →
Engineering

SQLite Is the Production Database You Already Know (You Just Don't Know It Yet)

March 5, 2026 · 9 min read
Automation

Building a WhatsApp Bot That Doesn't Get Banned — The Desktop Agent Approach

January 29, 2026 · 10 min read
Healthcare

Deploying a Full Radiology System on the Edge: 13 Modules, Zero Servers

January 15, 2026 · 12 min read

Browse by Topic

Explore articles by area of expertise

Use Cases

Real-world scenarios and how we tackle them

4 articles

AI & LLMs

Multi-provider AI, agents, prompt engineering

6 articles

Automation

Workflows, WhatsApp bots, voice agents

1 articles

Results & ROI

Measurable outcomes from our implementations

1 articles

Engineering

Architecture decisions, edge computing, performance

7 articles

Healthcare

Digital health, DICOM, FHIR, clinical systems

3 articles

All Articles

Search, filter, and explore the full archive

Results & ROI
12 min read

How a 3-Person Team Handles 200+ Leads Per Month with AI

A travel agency was drowning in WhatsApp messages and Excel spreadsheets. We built a complete digital storefront with ML lead scoring, AI chat, automated workflows, and voice agents — all on Cloudflare at $0/month hosting. Now they handle 5x more leads without hiring.

GM
February 26, 2026
Read Article →
AI & LLMs
9 min read

Why We Use 7 AI Providers (Not Just One) — And How We Track Every Cent

Vendor lock-in is a trap. Here's how our AI Gateway routes 11,200+ calls/month between Gemini, GPT-4o, Claude, DeepSeek, Groq, and more — with automatic fallback, cost tracking to the cent, and a ~$184/month total AI bill across 7 providers.

GM
February 12, 2026
Read Article →
Automation
10 min read

Building a WhatsApp Bot That Doesn't Get Banned — The Desktop Agent Approach

The official WhatsApp API costs $0.05–0.15 per message and restricts what you can send. Unofficial bots get banned in weeks. We built a third way: an Electron desktop agent with Playwright, human-like behavior, and zero bans after 8+ months in production.

GM
January 29, 2026
Read Article →
Use Cases
11 min read

No-Code Workflows That Actually Work in Production — 7,000 Lines of Execution Engine

Most "no-code" tools break at the first real-world edge case. We built a visual workflow engine with 20+ node types, Canvas API at 60fps, Durable Objects for long-running execution, and step-by-step debugging. Here's how 7,073 lines of engine code make drag-and-drop actually production-grade.

GM
December 15, 2025
Read Article →
Use Cases
10 min read

From 4-Hour Response Time to Instant: How Our AI Voice Agents Make Real Phone Calls

Twilio for calls, Gemini Flash for real-time conversation, ElevenLabs for 15+ natural voices. We built AI agents that confirm appointments in 35 seconds, qualify leads with 3 questions, and switch between Spanish, English, and Catalan mid-call. Plus: God Mode lets humans supervise and intervene live.

GM
December 29, 2025
Read Article →
Healthcare
12 min read

Deploying a Full Radiology System on the Edge: 13 Modules, Zero Servers

We built a complete RIS — DICOM viewer, MedGemma AI diagnosis, HL7 FHIR, digital signatures, CASS billing — running 100% on Cloudflare. Rated 9/10 by an expert radiologist who's used Siemens, Philips and Epic. Here's the full technical story.

GM
January 15, 2026
Read Article →
Engineering
11 min read

Edge Computing: Why We Bet Everything on Cloudflare (And What $65/Month Gets You)

No servers, no containers, no Kubernetes. We run 14+ interconnected products on 9 Cloudflare products — Workers, D1, R2, Durable Objects, Pages, KV, Vectorize, Workers AI and WAF. $65/month for what would cost $400-600 on AWS. Here's the full architecture.

GM
August 25, 2025
Read Article →
AI & LLMs
11 min read

Synapse Studio: A 2D Virtual Office Where AI Agents Do the Real Work

We built a SimTower-style animated office where AI agents with multimodal capabilities — vision, image generation, web search, iterative image evolution — collaborate on real tasks. Zero dependencies, pure Vanilla JS, running on Cloudflare.

GM
September 8, 2025
Read Article →
AI & LLMs
12 min read

Perspectiva Studio: 19,000 Lines of Vanilla JS That Create Audiobooks, Blogs, and AI Coach Sessions

We built a full content creation engine — audiobooks with 15+ ElevenLabs voices, blog articles with AI-generated images from 5 providers, PDF documents, and real-time AI Coach sessions — all in zero-dependency Vanilla JS running on Cloudflare.

GM
September 22, 2025
Read Article →
Healthcare
10 min read

Heartbeat: A Proactive Health Engine That Thinks Before You Ask

We built a health engine that beats like a heart — periodically analyzing user context through AI-powered Deep Pulse analysis, User DNA profiles, crisis simulation, and visual evolution to generate proactive recommendations before users even ask.

GM
October 6, 2025
Read Article →
Use Cases
11 min read

NutriNen Baby: A Gamified Baby Nutrition App with a 33-Tool AI Chatbot and a Music Box

We built a mobile-first baby nutrition app with meal tracking, a virtual fridge (6 categories), an AI chatbot with 33 function-calling tools, a music box with 22 melodies generated via Web Audio API, WHO growth charts, and full gamification — all in 19,600 lines of zero-dependency Vanilla JS running as a native Android app via Capacitor.

GM
October 20, 2025
Read Article →
AI & LLMs
9 min read

VOID: A Space Opera Game Where Your AI Copilot Remembers Everything and Generates the Universe

We're building a space strategy game where "Phantom" — a Gemini-powered AI copilot on your bridge — has persistent memory, generates contextual images with FLUX, creates missions based on your decisions, and makes every playthrough unique through generative narrative.

GM
November 3, 2025
Read Article →
AI & LLMs
8 min read

Transcriptor: From a Meeting Recording to Structured Minutes, Psychological Analysis, and an AI Retrospective Chat

We built a tool that takes a meeting recording and outputs diarized transcription, formal minutes with action items, individual psychological analysis of each participant, a WhatsApp-style retrospective chat, AI-generated summary images, and TTS narration of conclusions — all in 4,500 lines of vanilla JS.

GM
November 17, 2025
Read Article →
AI & LLMs
7 min read

Video Studio: AI Video Generation Using Image-to-Video with Progressive Motion Tiers

We're designing a video generation system that starts from existing AI-generated images instead of creating video from scratch. Four motion tiers — from free CSS Ken Burns to $0.30 AI video clips — let you balance quality and cost. A Story Director AI plans the motion per scene, and ffmpeg composes the final output in any format.

GM
December 1, 2025
Read Article →
Use Cases
10 min read

Cadences: The AI-Native Project Management Platform We Built from Scratch

Cadences isn't another Asana clone with an AI chatbot bolted on. It's a full platform with 7+ AI providers, voice agents that make real phone calls, WhatsApp automation, IoT control, visual workflow engine, offline-first sync, and edge computing — all built on React 18 + Cloudflare Workers with 107+ database migrations.

GM
August 11, 2025
Read Article →
Engineering
9 min read

SQLite Is the Production Database You Already Know (You Just Don't Know It Yet)

DHH hit 30,000 concurrent writers on SQLite. We run 14+ products on Cloudflare D1 (SQLite at the edge) for $5/month. SQLite isn't the toy database you think it is — it's powering everything from 37signals' ONCE to our entire multi-tenant platform. Here's why the industry is converging on the world's most deployed database.

GM
March 5, 2026
Read Article →
Engineering
10 min read

Vanilla JS Is the Assembly Language of the Browser (And That's Why We Use It)

Gmail ships 20MB of JavaScript. Slack ships 55MB. Our cookie consent system? 4KB, zero dependencies. When you understand the low-level primitives — vanilla JS, pure CSS, raw DOM APIs — you don't need a framework to tell you what the browser already knows. But we're not framework haters: we use Astro for SSG and React islands where they genuinely help. The difference is that we choose our tools — they don't choose us.

GM
March 13, 2026
Read Article →
Healthcare
14 min read

From Floppy Disks to Edge Computing: 25 Years of Healthcare Data

In 1998, a technician drove between hospitals every morning carrying a floppy disk with the day's primary care visits. The hospital's database was a multidimensional key-value store called MUMPS, accessed via Telnet over coaxial Ethernet. Lab analyzers connected through 32-port RS-232 cards. Today we run healthcare AI on Cloudflare Workers at 2ms latency. This is the story of what changed — and what didn't.

GM
July 14, 2025
Read Article →
Engineering
16 min read

The Strangler Fig Pattern in Practice: How We Migrated a Hospital from MUMPS to SQL in 4 Years

Martin Fowler wrote about the Strangler Fig Pattern as a theory. We lived it — 8 Caché/MUMPS servers, 40TB of unstructured medical data, bidirectional CDC at 100ms latency, and 4 years of coexistence between the legacy HIS and the new Angular/SQL Server system. Here's the engineering playbook, with the real numbers, the real architecture, and the real tradeoffs nobody talks about.

GM
July 28, 2025
Read Article →
Engineering
12 min read

RS-232, Packet Radio, and a SCADA Written in Lisp: What Modern Devs Can Learn from the 90s

A water utility's SCADA running C, Lisp, and TCL/TK on HP-UX workstations. Hospital labs with 32-port RS-232 cards parsing byte streams from 25 analyzers. Technicians driving floppy disks between clinics because the modem was down again. These systems worked — reliably, for years, with zero npm installs. Here's what they got right that we've forgotten.

GM
June 16, 2025
Read Article →
Engineering
14 min read

Why Every Database Migration Expert Started with Key-Value: From MUMPS Globals to D1

MUMPS globals were key-value before key-value was cool. Caché was a document database before MongoDB existed. The Entity-Attribute-Value pattern at Werfen was NoSQL hiding inside SQL Server. After 25 years migrating databases, I've learned that every storage model is a key-value store with varying amounts of ceremony. Here's what that means for how you design systems today.

GM
June 30, 2025
Read Article →
Engineering
14 min read

Local-First Desktop Tools: Why We Moved 73K Lines Off the Cloud

Not everything belongs in the cloud. ML model training on private data, IoT device communication via MQTT and Serial, browser automation with anti-detection, and real-time audio processing — we built 5 Electron desktop tools totaling 73K lines that run locally, sync with the cloud, and handle workloads that would be impossible to run remotely.

GM
July 15, 2025
Read Article →

No articles found

Try a different search term or browse all categories.