Podcast Summary

Podcast: Scrum Master Toolbox Podcast: Agile storytelling from the trenches
Episode: "AI Assisted Coding: How Spending 4x More on Code Quality Doubled Development Speed"
Guest: Eduardo Ferro, Head of Engineering at Clarity AI
Host: Vasco Duarte, Agile Coach
Date: February 18, 2026

Episode Overview

This bonus episode, part of "AI Bonus Week," features Eduardo Ferro discussing the real-world impact of AI-assisted coding on software development. Drawing from personal experiments and nearly 30 years of experience, Eduardo explains how investing heavily in code quality—using AI to amplify best practices—can actually double delivery speed. The conversation explores "Vibe coding," practices for both prototypes and production code, risk management, and how AI is changing team roles and architecture decisions in software development.

Key Discussion Points & Insights

1. Defining "Vibe Coding" and AI-Assisted Coding

• What is Vibe Coding?
  Eduardo defines "Vibe coding" as a flow-driven, curiosity-based approach to software development where the AI helps steer the process. It contrasts with traditional, meticulous manual review, making it ideal for fast prototyping, experimentation, and MVP work.

  • Quote:

    "For me Vibe coding is flow driving, a curiosity-based way of building software with AI. And it's less about meticulously reviewing each line of code or more about letting the AI to steer the process." (Eduardo, 02:32)

• Distinguishing Contexts of AI-Assisted Coding:
  For production code, Eduardo emphasizes structured practices: detailed specs, edge case handling, and vertical slicing—contrasting the more exploratory, flexible nature of Vibe coding.

  • Quote:

    “It's not the same... developing like a one shot script... or like an MVP... or I am developing like production code... it’s about managing the risk.” (Eduardo, 04:04–06:20)

2. Key Practices that Work for AI and Humans Alike

• AI Amplifies Good Engineering Practices:
  Eduardo discovered that practices beneficial to human developers—tight feedback loops, small safe steps, clear domain-driven code, automated testing, and frequent refactoring—are equally potent when interfacing with AI.

  • Quote:

    “What already works for humans works very well for AI... All of these best practices... work super well also for AI.” (Eduardo, 07:25)

• Investment in Code Quality is Cheaper and Faster:
  AI enables more frequent and thorough code review, testing, and refactoring—at a fraction of the previous effort and time.

  • Case Example:
    Eduardo invested four times more effort into code quality tasks but doubled his delivery pace due to AI assistance.

    • Quote:

      "I was investing four times more in refactoring, cleanup, deleting code, introducing new tests, improving the testability, security analysis than in generating new features... At the same time globally I... double my pace of work..." (Eduardo, 08:22)

  • Feedback Loop Acceleration:
    “I have like a small agent that I put in an automatic way... It analyzes the code, suggests refactoring, and can even generate merge requests. So in the morning, I’d have two or three improvements waiting for me.” (Eduardo, 09:30)

3. Why Removing Code Is a Feature, Not a Bug

• AI Makes Clean-up and Architectural Change Easier:
  Removing unused code is easier, which, in turn, makes refactoring and architecture changes faster—producing a positive spiral.

  • Quote:

    “AI can also make removing code faster. Removing code makes changing architecture easier... as we go, we spend time effectively removing code... which is legacy, but is also liability.” (Vasco, 13:18)

  • Amplification Effect:

    “If you have bad practices, the collapse is super fast. If you have good practices, I think that you can even accelerate.” (Eduardo, 14:27)

4. AI as an Amplifier: The Double-Edged Sword

• Good Engineering Gets Better; Bad Gets Worse:
  AI amplifies existing practices. Well-organized teams with robust processes benefit the most, while those with poor practices risk faster failure.

  • Quote:

    “AI is an amplifier. If you have bad practices, the collapse is super fast. If you have good practices, you can even accelerate.” (Eduardo, 14:27) “...the people who already knew how to develop software well will continue to develop it well and faster. And...those who did not...will probably get into troubles much faster...” (Vasco, 19:25)

• AI Doesn’t Replace Good Judgment or Architecture:
  Mistakenly treating AI as a magic bullet, or ignoring foundational engineering, can yield fragile systems and quick failures.

  • “There's push from companies to introduce AI without proper preparation. If you introduce it without feedback loops, or without having tests, you can have a problem.” (Eduardo, 16:55)

5. Shifting Bottlenecks and Evolving Roles

• From Coding to Decision-Making:
  As implementation gets faster, bottlenecks shift to product decisions, market validation, and discovery.

  • "If coding or implementing is not the bottleneck, we will need to see how they will evolve...our profession." (Eduardo, 24:31)

• Collapse of Traditional Roles:
  Developers may find themselves acting as strategists, product managers, designers, testers, and coders all at once, especially in smaller teams.

  • Quote:

    “Right now a developer is a strategist, a product manager, designer, a coder, a tester... roles start collapsing into one person.” (Vasco, 25:17) “I expect super small teams in more fluid organization... teams of three, four, but not six, seven or so.” (Eduardo, 27:14)

6. Architecture Decisions Become More Critical—And Frequent

• Risks of Fast, Easy Choices:
  With AI making everything rapid and frictionless, there’s a greater risk of making the wrong architectural or technology choices, which can have lasting effects.

  • Quote:

    “It's very quick...to make bad architecture decisions, but it's also very easy to make bad technology use decisions.” (Vasco, 20:34) "You will need to do more architecture decisions per week or per month... but you need to have the taste to detect that the architecture is the problem..." (Eduardo, 21:54)

• Possible Commoditization:
  Common frameworks and platforms may become even more commoditized since AI tools are trained more on the dominant solutions.

7. Trends and Future Outlook

• Tools Will Absorb Best Practices:
  AI-assisted and augmented coding practices will become embedded in development tools, lessening the manual burden of enforcing best practices.

  • “Some of these practices will be incorporated even inside the tool...In general, I think that we are going in the right directions.” (Eduardo, 23:43)

• Role Fluidity & Team Size:
  Teams will become smaller, more mission-oriented, and fluid in their composition, with AI acting as a thinking partner to fill gaps in knowledge or expertise.

• Marketplace Saturation:
  The speed of AI-assisted development could lead to a flood of new software, making discovery and distribution channels even more vital.

Memorable Quotes & Timestamps

• “Vibe coding is flow driving, a curiosity-based way of building software with AI...”
  — Eduardo Ferro [02:32]

• “What already works for humans works very well for AI.”
  — Eduardo Ferro [07:25]

• “I was investing four times more in refactoring, cleanup ... but ... double my pace of work.”
  — Eduardo Ferro [08:22]

• “AI is an amplifier. If you have bad practices, the collapse is super fast. If you have good practices... you can even accelerate.”
  — Eduardo Ferro [14:27]

• “Right now a developer is a strategist, a product manager, designer, a coder, a tester... roles start collapsing into one person.”
  — Vasco Duarte [25:17]

Important Segment Timestamps

• [02:32] – Eduardo defines "Vibe coding" and distinguishes it from other AI-assisted coding.
• [07:25] – What works for humans works for AI: best engineering practices amplified.
• [08:22] – Eduardo details his experiment: 4x effort on quality, double speed.
• [13:18] – The value of removing code and fast architecture changes using AI.
• [14:27] – AI as an amplifier: makes good better, bad worse.
• [16:55] – Risks when organizations adopt AI without preparation.
• [19:25] – AI increases the capability gap; reinforces best/worst practices.
• [20:34] – The dangers of easy, fast technology and architecture decisions.
• [23:43] – Eduardo's perspective on future trends for AI-assisted code and teams.
• [25:17] – The collapse and fusion of roles in software teams.

Resource Recommendations

• Eduardo’s blog & recommendations: efero.net (Contains talks, latest experiments, recommended channels and books)
• Kemp Farley modern software engineering YouTube channel – For AI & XP insights
• Book: "Vibe Coding" by Jin Kim et al. (noted as evolving rapidly)
• General advice: Use your AI assistant to help discover more up-to-date resources, and experiment.

Summary for Non-Listeners

This episode unpacks how AI can act as both a productivity amplifier and a risk multiplier in software engineering. Eduardo Ferro’s hands-on experiments show that focusing more on code quality and review—with significant AI assistance—not only keeps codebases healthier but also dramatically increases delivery speed. Yet, both Eduardo and host Vasco highlight the need for engineering discipline, the ability to make sound architectural decisions quickly, and an evolving definition of software roles. As AI pushes coding speed and team fluidity to new heights, the real bottlenecks shift upstream to product discovery, decision-making, and market validation.

For practitioners, the takeaways are actionable: double down on engineering best practices, think carefully before making architectural choices, expect smaller, more mission-focused teams, and embrace continuous learning as AI tools and methodologies rapidly evolve.