Harnessing AI to go beyond vibe coding: Insights from our Sydney G[=]K talk

Is vibe coding the future of software delivery or a risky way to build products by swapping rigorous engineering discipline for speed and agility? Or is the reality somewhere in the middle?

At our latest G[=]K event in Sydney, experts unpacked why AI-driven “coding by vibe” offers exciting opportunities for speedy software delivery, but why it must be balanced with responsible, secure reviews and testing processes.

Titled: “Supercharging software teams: Harnessing AI to go beyond vibe coding,” the event featured not one but two expert speakers from the Equal Experts Network in Australia, providing a dual perspective on AI within software delivery and product development.

Despite the poor weather, including a spectacular spring thunderstorm, it was standing room only at the event. Those who braved the rain were rewarded with a chance to gain insights on AI-driven software delivery and share their own vibe-coding experiences with others over great food and drinks.

Vibe coding is good for entrepreneurs wanting a quick trial, but not so much for enterprises

Senior Technical Associate Dominik Fretz kicked off the talk with an introduction to vibe coding and how AI is enabling more people, not just software developers, to generate code for their own ideas.

He said: “Vibe coding might not be the best for all applications. For big corporations with giant code bases and years of development, it is probably a bad idea to let it loose on a project like that. But where I see vibe coding excel is as a quick shot, an experiment as a way for entrepreneurs to try something out and see what happens, see if it is worth developing further.”

Dominik explored the proliferation of AI coding tools and how new ones are constantly being created, with a vibe coder’s context and purpose playing a vital role in determining which tool is best for their needs.

But how can software delivery teams use AI to improve? Dominik argued that it’s all about context. He said: “Context is king. Give it as much information, treat it as a junior developer that is very fast typing, but you have to tell them what to do and how to do it, otherwise it will just go into something else.”

He also shared practical implementation strategies. For teams just starting out that might be starting small and learning what AI tools can do before graduating to more detailed products, all the time looking to establish best practices. For more mature teams, it can involve implementing spec-driven workflows, creating project-specific steering files and establishing agent hooks and context engineering standards to make AI development more effective.

He also shared best practices to help go beyond vibe-coding, including:

• Review everything: Never commit AI code without understanding it. Just like humans, AI can make mistakes, and it can hallucinate.
• Test rigorously: AI is great at making code that looks right but is, in fact, wrong.
• Document intent: Capture the “why”, not just the “what”.
• Measure and validate: Track quality metrics, not just speed.
• See what’s new: This field develops quickly, so update processes where applicable.

“Vibe coding is great for a demo or a quick trial, but you then need to use structured approaches to ensure quality. Don’t blindly trust the AI code. It does make mistakes. Always review, test and approve.” Dominik concluded.

Uplifting an IoT platform in one month with AI

Seb Jacobs, Senior Business and Product Associate, then shared his experiences from a product perspective, taking the audience through his journey creating his business Conexie.

After starting with an idea for a platform enabling people to report problems that need fixing, Seb eventually pivoted to create an end-to-end IoT solution providing actionable insights. And it really took off. But sadly, the platform wasn’t ready to scale. So Seb turned to an AI agent to help move the product forward.

He explained: “You can’t just use AI to build a platform and have it go straight to market. What we wanted to do was have something that I could take to my team of developers and say – This is exactly what I want. Can you make this work better, in a more robust environment?”

What had previously taken Seb and his team five years to create was now reduced to just one month of work, at significantly less cost and with the added ability to immediately input and test new ideas.

He said: “A lot of people say AI is only good for prototyping, but if that’s all it’s good for, why not use it? Prototyping is a really clunky part of the product process. With AI, you can actually sit with a client and demonstrate something and change it in front of them.”

Seb concluded with some warnings about the grandiose nature of some AI agent tools, and the reality being that AI can assist in the process, but humans are still vital for their prompting skills, reviewing the AI-generated code, testing and iterating.

“AI agents are accelerators, not replacements. Use it for speed and agility, but engineering discipline is always required. And remember, no matter what you say, the AI Agent will always reply, “Absolutely, that’s a great idea”. It will not always be a great idea.”

Join our growing technology community in Australia

Our G[=]K, also known as G-EE-K, technology event series officially launched in Australia last year. But these events are more than just technology talks as they embody our core values of knowledge sharing, fostering communities and uplifting teams with new skills or capabilities.

By bringing together like-minded technologists at events like G[=]K, we aim to share the knowledge and skills within our Equal Experts Network and inspire attendees with new ideas or approaches that they can take back to their own organisations.

Missed out on this event? Watch out for details of future G[=]K Australia talks in Sydney and Melbourne. If you’re interested in learning more about the Equal Experts or our Network, contact me and the rest of the Australia team now.

This blog reflects the personal experiences and opinions of the author and the G[=]K event speakers. It should not be taken as formal Equal Experts methodology.

About the author

Prasanna Murugavel is Head of People and Community at Equal Experts APAC. With more than 17 years of experience in global HR, Prasanna is a hands-on people leader committed to achieving organisational success and accelerated business growth. Connect with Prasanna on LinkedIn.