Loop Engineering: Automation in the Age of AI
Everyone's saying we should write loops, not prompts. Strip away the novelty and it's automation — with one new thing sitting inside the loop's body.
CTO & Co-founder at DailyBot (YC S21). Full Stack Developer, speaker, and tech entrepreneur with 14+ years building digital products from Colombia.
Everyone's saying we should write loops, not prompts. Strip away the novelty and it's automation — with one new thing sitting inside the loop's body.
An agent nailed the first hour and drifted by the third. The fix wasn't a smarter model — it was a durable plan the repository itself could hold the work to.
Midjourney, the AI image company, unveiled a full-body ultrasound scanner and a wellness spa. Is it a pivot or a new division — and does the tech actually work?
Harness engineering is the new way to steer AI agents — the system around the model that checks its work and stops it repeating the same mistake.
Anthropic said a model was too dangerous to release. Then they released it — same weights, new name, and a safety layer that changes the equation.
An agent that can install a skill from a URL stops reinventing the wheel. A field report on the standard agents are quietly building beneath the surface.
Public package registries have shipped malware over and over this past year. What's happening, why AI changed the game, and how to lock down the install side.
How I built a two-type slide deck catalog inside my Astro website — discriminated unions, AEO twins, asset isolation, and live theme sync.
A hands-on comparison of Reveal.js, Slidev, Marp, Spectacle, and more — with a feature matrix to pick the right slides-as-code tool for developer talks.
Five categories, eight artifacts, one concrete guide: what your site needs to be agent-ready — including the commerce layer I deliberately don't fake.
State of the art of mobile development in 2026: the available options, how each one works, and where it actually makes sense to start today.
My FLISOL talk at UTP Pereira: the story of OpenClaw and its creator Peter Steinberger, and how a one-hour WhatsApp bot reshaped the AI-agent world.
Around 30 launches in one week. Everything Cloudflare shipped at Agents Week 2026 — runtime, MCP, standards, measurement — organized by layer.
My Abril Mobile talk at Universidad Católica de Pereira: how a full-stack dev chose a path to learn mobile in 2026 — and why I picked Kotlin Multiplatform.
The story of OpenClaw and the birth of personal AI assistants — a new paradigm of local, open agents owned by you, not by a company.
The Moltys meetup at the UTP in Pereira — an OpenClaw community workshop with Cursor where I walk through its history, use cases, and architecture.
A journey through the life and work of Isaac Asimov, from the Foundation saga to the Three Laws of Robotics, and why his predictions are coming true in the age of artificial intelligence.
A sci-fi reading club is born inside the Pereira Tech Talks community. In this series I share reviews and reflections on the stories, movies, and tales we read and watch each week.
Anthropic built a model that found thousands of zero-days in every major OS. They won't release it — they assembled a $100M coalition instead.
Most people think an agent is an LLM with tools. They're wrong. A real agent is a system — and building one requires a different kind of engineering.
I didn't get the hype around a text measurement library. So I built dozens of demos — from masonry grids to a sphere that reflows text at 60fps.
AI agents now have payment cards, wallets, email, and social networks. A new economy is forming — and it's not what I predicted in 2018.
AI was supposed to give us more free time. Instead, it created an addictive building loop — backed by neuroscience, economics, and a meetup confession.
AI agents aren't productivity multipliers — they're judgment multipliers. Why direction matters more than the model, and what that means for engineering.
How I went from writing every line of code to orchestrating AI agents — and why the biggest tech revolution is happening so quietly that almost nobody sees.
How I migrated XergioAleX.com to Astro 6 in minutes, what broke, and why the experimental Rust compiler turned perception into measurable speed.
AEO measurement is years behind AEO optimization. Here's what the industry has, what's still missing, and one audit framework that gives you something concrete to work with.
Markdown for Agents: content negotiation meets AI—what the pattern means and a working Astro plus Cloudflare Pages implementation.
The search landscape changed — and most sites aren't ready. Here's what the shift looks like, why AI answer engines work differently, and the two foundations every site should have in place.
AI agents are now some of the most important visitors to your site — and they don't click links. A practical guide to Answer Engine Optimization: what it is, how it differs from SEO, and why the first-mover window is still open.
After years of chasing ESLint upgrades and managing the Prettier conflict dance, I switched to Biome. One config file, one binary, and I haven't looked back.
How building a cargo transport platform with university friends taught me the most valuable entrepreneurial skill: telling a story that makes people believe.
One Cloudflare Pages middleware file reveals AI crawler visits that client-side JavaScript analytics completely miss—no extra scripts required.
After years deploying static sites on GitHub Pages, I evaluated the free options and landed on Cloudflare Pages. Here's what I learned and what surprised me.
Designing multilingual Astro from scratch: typed translations, the Page Wrapper pattern, and scaling to more languages without rewiring components.
Why Astro and Svelte simplify modern web development—State of JS 2025 data, benchmarks, and practical lessons from building this website.
How this blog scales without a backend: Content Collections, tag taxonomy, client search, series navigation, and bilingual builds.
Chapter four while building XergioAleX.com: free analytics with privacy-first tools that keep Lighthouse at 100—no cookie banners.
The story of reaching 100/100/100/100 on Lighthouse PageSpeed for Performance, Accessibility, Best Practices, and SEO on mobile and desktop.
From one-page landing to bilingual, AI-ready XergioAleX.com: architecture, stack choices, and why Astro + Svelte fits static sites.
Pereira Tech Talks with Sebastián Mora — from writing code to orchestrating agents, what MCPs unlock, and a robot at a software meetup.
From my Vibe Coding e IA workshop at Open Tech Hackathon — Cursor vs Codex vs Claude Code, conductor mindset, shipping AI at DailyBot.
Astro and Svelte portfolio site for Sebastián Martínez Vanegas, Colombian poet and winner of the Emilio Prados International Poetry Prize.
A fully static wedding invitation site built with Astro, Svelte, and GitHub Pages — unique codes, RSVP via Google Forms, zero backend.
How I moved from manual MetaTrader 4 trading to migrating scripts, indicators, and expert advisors into MetaTrader 5 for stronger automation.
Migrating pereiratechtalks.org from Dockerized Ghost to static Astro on GitHub Pages — ten years of lessons packed into two weeks of code.
Why Astro won me over — Microsoft and Firebase cases, measurable gains, and a live Pereira Tech Talks migration to prove it is not hype.
Market Profile as a visual framework for imbalance zones, institutional flow, and structured trading decisions instead of pure intuition.
From futures with friends to Forex and Market Profile — discipline, psychology, and consistency matter more than chasing fast profits.
What changed between May and August 2023: Manizales Tech Talks on AI transforming app development, three months deeper into the revolution.