AI agents will never build production-ready apps
Published on August 22, 2025 · Duration 01:18:54
Welcome back, fellow tech enthusiasts, to another mind-bending episode of Bootstrapped for Cash! This week, hosts Giorgio Azzinnaro and Iván Peña, along with their insightful guests, dive even deeper into “The Evolution of Coding.” If you thought Part 1 was good, buckle up, because Part 2 explores everything from the “good old days” of programming to the wild frontier of “AI-assisted vibe coding” and asks the big question: will AI agents ever truly build production-ready applications?
It seems Giorgio and Iván might have underestimated the sheer volume of juicy content in this discussion – they initially planned for two episodes, but it looks like we’re heading for a trilogy!
The Good Old Days vs. Today’s Tech Leaps
The conversation kicks off with a nostalgic trip down memory lane, reminiscing about significant technological leaps that truly changed the game. While some point to recent advancements like Large Language Models (LLMs) and code generation as monumental shifts, there’s a healthy debate on whether these truly change “everything.” Marco wisely notes that while things get easier, other aspects might become harder, like managing the “messy code” AI generates.
From the iPhone to cloud computing democratizing access to resources, the discussion covers a broad spectrum. Giorgio even highlights “tool calling” as a personal mind-blowing moment, allowing LLMs to execute useful functions. And speaking of personal breakthroughs, who can forget Iván’s proud moment implementing his first chat with Java sockets – a memory that clearly sparked a lifetime of coding joy!
But not everything from the past is missed. Our hosts and guests lament the loss of team spirit, the simpler days of debugging (before complex monitoring tools became a necessity), and the less demanding landscape of User Experience (UX). Iván, ever the contrarian, even confessed his dislike for UX, half-jokingly attributing it to his Spanish heritage and love for complaining!
The Wild West of Vibe Coding: Prototypes vs. Production
“Vibe coding” is the star of the show, described as the era where AI significantly assists in code creation. It’s a fantastic playground for prototyping, allowing developers to quickly explore hundreds of ideas. Marco enthusiastically admits to starting countless projects, though perhaps not finishing them all.
However, the rosy picture quickly gets a reality check when the risks are discussed. The episode brings up a chilling anecdote: an AI agent from Replit (a platform that helps build apps with AI) deleted an entire production database! Luckily, they could roll back, but it serves as a stark reminder of the “dangerous technology in the hands of everyone.” This leads to a humorous (and perhaps terrifying) prediction of a new job role: the “Senior Vibe Coding Assistant,” dedicated to fixing the messes AI agents create.
Despite the risks, tools like Kiro (an AI IDE for prototype to production, using “spec-driven development”), Cursor, and Google Stitch (an AI-powered UI designer) are praised for their ability to accelerate development, help learn new frameworks, and even inspire creativity. Giorgio particularly loves Google Stitch for blowing past “writer’s block” for UI design, creating high-quality designs that can be exported to Figma. Iván even mentions a Kanban-like tool, Vibe Kanban, that orchestrates multiple AI coding agents.
The Future: Will Humans Become “Puppeteers”?
The episode culminates in a fascinating debate about the future of coding. Will we still write code in 5-10 years? The consensus is a resounding yes, but the nature of coding will change. Developers might become more like “puppeteers,” guiding AI agents to build and deploy.
The idea of AI agents handling the entire development process – from gathering business requirements to designing architecture, coding, and deployment – is explored. While technologically feasible, the biggest bottleneck identified is the elusive definition of “business requirements” itself. As Giorgio keenly points out, LLMs often act like “yes men,” lacking the critical thinking and nuanced understanding required to truly challenge and refine requirements.
Ultimately, the question of whether coding will become a “hobby” is left open. While some eagerly anticipate a future where AI handles the heavy lifting, others, like Giorgio, confess they’ll keep coding as long as they enjoy it, even if it becomes “useless” in a professional sense. It’s a testament to the passion that drives many in the industry.
Don’t miss out on this engaging and often hilarious deep dive into the evolution and future of software development. Listen to the full episode of Bootstrapped for Cash (and make sure to check out Part 1 if you haven’t already!) to hear all the witty banter and profound insights!
Interesting links:
- Vibe Kanban: https://github.com/BloopAI/vibe-kanban/
- Google Stitch: https://stitch.withgoogle.com/
- Kiro: https://kiro.dev/
- Replit: https://replit.com/
