From Vibe Coding to Scalable, Reviewable and Performant Engineering
About this event
We are back with two great speakers!
💡 First on stage is Bart Waardenburg, the creator of the tool Fallow. His tool gained significant traction after it was discussed on the web development podcast Syntax.
Vibe Coding Doesn't Scale: Deterministic Tooling for Agentic Engineering
AI Slop Is a Tooling Problem
Most AI coding fails in the same way. The model writes plausible code, but it has no deterministic way to check whether the change made the codebase better or worse. We compensate with prompts like "make no mistakes", "follow best practices", or "act like a senior engineer".
In his talk, Bart shows where the line runs between AI slop and real agentic engineering. Using Fallow as the case study, he walks through how an agent calls deterministic codebase-quality checks through CLI, MCP, and agent hooks, so engineering standards live in the loop itself instead of in wishful prompts. The agent generates code, inspects the repository, detects dead code, duplication, complexity hotspots, and boundary violations, optionally combines that with runtime evidence, and uses those signals to refactor or self-correct before a human reviews the diff.
The core idea is simple. The biggest productivity gains do not come from telling the model to be smarter. They come from giving it a better harness. For frontend teams, that is the difference between vibe coding and scalable, reviewable engineering.
💡 After the break, we welcome Sander van Surksum on stage. Sander is a web performance specialist who works with global brands to build fast, stable and world-class digital experiences.
Vibe coding made the web sick. Now it can fix it.
Since ChatGPT went live in November 2022, the median mobile web page has been getting heavier three times as fast as before. AI didn't invent bloated front-end code; it learned from a "lost decade" of bad patterns and industrialized them.
Sander follows the evidence: the public HTTPArchive data that marks the inflection point, and Alex Russell's Performance Inequality Gap, which asks the question most teams never do — whose device are we actually testing on? Spoiler: the bill isn't paid by the developer on a MacBook Pro and fibre. It's paid by the user on a €300 Android, 9 Mbps. The same code, two completely different realities.
Today's median page is already bigger than the entire game DOOM.
The talk ends with good news. The same AI that made the mess can help clean it up, once you give it eyes. We'll look at the 2025–2026 tooling (Chrome DevTools MCP, Google's Modern Web Guidance) that finally lets AI agents measure real performance and ship lighter code, and close on the one question every engineer should answer before their next prompt: who are you building for?
📍 Location
We welcome you to our office in Utrecht: Studio N in the Werkspoorkathedraal. The address is Tractieweg 41. We have plenty of free parking space available. We will have dinner ready for you at 18.00!
🕔 Agenda
18:00 Walk in & food
18:30 First talk
19:30 Break
20:00 Second talk
21:00 Drinks & Bites
Both talks are in English. RSVP now for an evening of learning and fun.
Don't miss out on this amazing opportunity to learn and connect with our community of frontend developers!
Source: meetup