Oredev Conference 2026 Part I
This is Part 1 of a 2 part series where I share my highlights from some of the talks I attended at the 2026 Øredev Developers conference.
2026 was my second year being a co-chair for the Quality and Test track alongside Maria Kedemo.
We aimed to curate a fantastic track for attendees who are interested in testing and quality, after working together on this for the 2025 Oredev conference.
Agentic AI in the Browser: Beyond Chatbots
Lucky Nkosi shared his experiences in working with Agentic AI in the browser and how it can make decisions, take actions etc. He also talked about the importance of making sure the Agents don’t get in the way of human users from taking actions.
He shared how in Africa, you have to deal with the limitation of poor infrastructure/ very expensive data. He also shared how in South Africa, Call centre agents at the tax centre face a particularly interesting challenge as they need to be able to support 11 languages, to help people in 11 languages.
An LLM he talked in more depth about in his talk was WebLLM - an in-browser LLM inference engine that brings language model inference directly onto web browsers with hardware acceleration. Everything runs inside the browser with no server support.
Fun fact: Learned that 72% of Enterprise AI spend goes towards LLM solutions.
Useful Resources from Nkosi’s talk
Mapping Biases to Testing in the LLM era
Maaike Brinkhof gave a more modern version of an earlier talk she gave on mapping biases to testing, with a focus on LLM usage, given how prevalent it is in the industry today.
She emphasised the fact that unbiased doesn’t exist. Everyone and everything (LLMs) are biased because we have lived experiences, which shape us.
Brinkhof went over the differences between System 1 and System 2 thinking, the benefits and costs of both and some examples of cognitive biases in testing. She also shared the unofficial working agreement she has with development teams when she joins them as a tester.
There’s a big problem in testing, and it’s the fact it is misunderstood.
With testing, you cannot prove the absence of something (bugs), you cannot guarantee quality.
Testing is the search for (new) information (that matters).
Brinkhof then addressed the fact that now, testers are now testing in quicksand - work is now on such a fast pace.
The increasing use of LLM/AI is leading to peopole caring less about their work. When you use a tool, you are trusting the tool, even when you don’t know what goes on beneath the hood.
Despite the fact it is good to slow down to build better software, everything wants us to speed up.
Useful Resources from Brinkhof’s talk
Practical Guide to Agent Native Platforms
Jessica Andersson gave a talk on building developer platforms where your users are both humans and Agents.
Andersson reminded us that now development is moving at such a fast pace, we are now moving from writing code to orchestrating change. (In one of her slides, she shared that according to public GitHub repositories, there are about 3-4x the number of issues created and Pull requests merged in Q3 2026, compared to Q3 2025).
She stated that Internal developer platforms are the foundation for succeeding in the AI era. Platforms are built to serve common needs in an organisation once. If the platform team make changes, they need to make sure it’s easy for the teams (they support) to go along with these changes. Andersson shared how important and helpful it is to have a Golden path, as any “small” change/different way of working done by one team isn’t actually a small exception: the fragmentation has an immense blast radius.
It’s expensive to have a too-fragmented tech stack.
Useful Resources from Andersson’s talk
## Three Disasters That Started With “We Don’t Need to Test That”
Christine Pinto gave an interactive talk where she shared 3 examples of IT disasters.
Before going into each example in more depth, she asked the audience, what did they think was the cause of the problem.
Everything the Tester in Your Life Wants You to Know About AI (But Hasn’t Figured Out Yet)
Vernon Richards gave a talk on the new reality testers face as they work in software development, with the increasing use of AI.
Richards addresses the identity crisis testers are starting to face, the evolving landscape and how testers and the teams around them can adjust.
Jobs aren’t disappearing, but tasks are being repriced. The role of the tester needs to evolve to meet this changing landscape.
Richards reminds us that if you can’t describe the work, AI can’t do it. Some tasks that testers do, are easier to replace with AI than others. He also encourages to be intentional with how testers talk about their testing and the value they add to the team.
Making room for testing - empowering teams to deliver business value with confidence
Andrei Contan shares his experience in using metrics when looking into quality and being mindful of how teams add business value, during his time at Booking.com. He also shared the importance of doing experiments and actually seeing if it was successful- now just hoping/waiting for it to be a success.
Contan shared the unique challenge he faces at his employer where there are so many versions of their website, that each refresh leads to a different (version of their) page.
“The goal wasn’t to be right, it was to learn fast.”
However, AI makes things too easy. There is now less friction - this adversely affects the learning journey.
Quality practices work when they create fast, safe, actionable feedback. (We don’t want judgement after the fact).
