Skip to content

Engineering leadership for the agentic era

More code than ever.The bar stays where it was.

I'm Felipe Plets, a seasoned engineering leader. Vibe coding and agentic engineering changed how fast software can be produced, but you still need a good strategy for high-quality code. I work with teams on the practices that let velocity rise without sacrificing quality.

The strategic question for engineering leaders isn’t “how do we generate more code?” but “how do we design our engineering practices so quality doesn’t drift as velocity increases?”

40%
of development effort goes to avoidable rework and maintenance driven by technical debtMcKinsey, 2023
$2.41T
annual cost of poor software quality to the US economy, with ~$1.52T in accumulated technical debtCISQ, 2022
~3x
more commits pushed to GitHub in a single month of 2026 than in all of 2025GitHub, 2026

About

Since 1999, building, leading, and reimagining software engineering.

Where it started

I started programming in 1999, pulled in by what still makes software worth doing: an idea can become something other people rely on, and what separates the two is mostly craft. Since then I have shipped production systems in Delphi, PHP, C#, Swift, Kotlin, Python, Ruby, ABAP, Java and, still, a great deal of JavaScript. Along the way I published academic work on ubiquitous computing, including a collaborative telemedicine system built to improve diagnosis quality where doctors were scarce.

Leading teams

I have been in leadership roles since 2006: developer, team lead, coach, consultant, director, founder. Different titles, the same job. Helping teams design a development lifecycle that produces consistent gains in productivity, quality, performance and security, then holding the line on it. I have done that with teams in different countries and in every arrangement you can name, from co-located rooms to groups spread across time zones to fully remote organizations.

What I focus on now

Today I work as a director of engineering and write about the shift happening around all of us. The promise of AI-assisted and agentic development is real, and so is the failure mode: organizations that get dramatically faster at producing code they no longer understand. My work is about taking the first outcome without the second, and most weeks I am comparing notes with other engineering leaders living exactly that problem.

The practice

Four things I keep coming back to

AI changed how quickly code appears. It did not change what makes software worth trusting. Most of my work sits at that intersection.

01

Quality that does not drift

Vibe coding and agentic engineering dramatically increase how fast code can be produced. High-quality code still requires planning, architecture and experienced judgement. Without strong guardrails, average quality drifts down quietly, and then all at once.

  • Guardrails by design
  • Architecture up front
  • Risk in the dependency chain
02

AI-native teams

The biggest leadership challenge right now is not AI adoption. It is building teams that collaborate with AI rather than consume its output the way they once consumed search results.

  • Collaborate, not consume
  • Redefine "good work"
  • AI fluency as a norm
03

Alignment, human and machine

Alignment has always been the hard problem in organizations. As AI becomes part of the work, that alignment has to extend to how teams use models: which ones, for which tasks, judged against which standards.

  • Shared standards
  • Context and instruction
  • Review against agreement
04

Evidence over opinion

Leaders need to talk about AI with evidence, not vibes. Models change weekly and the claims about them change faster. Decisions should rest on how tools perform on your work, not on how they are marketed.

  • Benchmarks and telemetry
  • Measure the outcome
  • Revisit as models move

Writing

All You Can Lead

Leadership, technology, and personal growth, written from the lens of a hands-on engineering leader.

Get it in your inbox

Essays on leadership, AI-assisted delivery and the craft of building software that lasts. No spam, unsubscribe anytime.

Subscribe free

Latest essays

All posts

Speaking & work

Ways we can work together

Most of my engagements start with the same conversation: the team is using AI, output is up, and nobody is quite sure whether that is good news yet.

Keynotes & talks

Conference and internal sessions on what actually changes when agents join the team, and what stubbornly does not.

Team workshops

Hands-on sessions that move a team from ad-hoc AI usage to a repeatable, reviewable way of building.

Advisory

Working with engineering leaders on delivery practices, quality strategy and the operating model behind them, informed by having built companies as well as run engineering inside them.

Felipe Plets speaking during a radio interview

Contact

Let’s talk about what your team is actually shipping.

Whether you are three months into an AI rollout or still deciding where to start, I am happy to compare notes. The newsletter is the easiest place to keep in touch, and everything else is below.

Subscribe to All You Can Lead