A model I was working with scored R squared 0.998 on public US data. One of the input columns had been calculated from the target by the same agency that published both, so the model was rediscovering a formula it had already been handed. Tests don't catch this. Nothing throws, nothing fails, the numbers look great. So I built a linter for it. You write a manifest of what derives from what, the tool walks the graph, and CI goes red when a feature sits on a derivation path to or from your target. It returns four verdicts, including one for columns whose provenance nobody has traced yet. The code is Python. The idea applies anywhere you have a dependency graph and a CI job. I'll show the leak, the manifest, the linter, and a run going red live.
Repo: https://github.com/Adeniyikayodee/derives-from-tutorial

At 15, I’m using AI across research, learning, business building, content creation, decision-making and automation. This session will break down the practical systems I use, what I automate, what I keep human, and what I’ve learned from experimenting with AI in real work.
How to use AI to learn, research and understand information faster.
How to identify repetitive work that can be turned into practical AI workflows.
How to combine AI with human judgment to build better systems for work, business and productivity.

"To love without understanding harms the object of your love" (Thich Nhat Hanh) For us, to code without understanding harms the product of our code -- our software. And quietly hollow out our own skills in the process. So how can you use AI as your coach, your teacher, and your partner in exploration? How can you prevent the brain rot of passivity? And when you're not writing it -- how can you keep all of the new code in your head? This talk aims to ensure your favourite LLM makes you a better developer, not a worse one. I'll share a system for collaborating on software that leaves you sharper, more skilful, and still the one who understands what's running in production.

What exactly is copyright? What about a licence?

Have you ever wondered if the licence of the package you installed from npm after some Googling (or AI-ing) affects you and your code in any way?

And does anything change in this new AI era?

About Alexandru: Alexandru works at BENlabs by day and is an amateur game developer and musician by night. In a past life, he also co-founded Drupal Romania and spent years promoting and teaching Drupal. From time to time, he also likes to ride pig statues while pretending to be Bastian from The NeverEnding Story.

Simple text-based prompts and chat interfaces are the most common ways of interacting with LLMs. But when it comes to image generation models like Stable Diffusion, a simple text interface doesn't provide users with enough creative control over the output.

This talk will demonstrate an approach we took at Stability AI to solving this problem that combines AI prompt enhancement, auto-model routing, mood boards, and full human-in-the-loop creative control.

I'll cover how we came to the UX decisions we made, how we built this experience on top of the Cloudflare Agents SDK, and some of the difficulties we ran into during implementation.

Finally, we'll talk about the impact these UX changes had on our users and our product. Was it a disaster? Come to the talk to find out!

About Jamund: Jamund lives in London with his family and dog. He currently works at Axle Energy on smart EV charging. He loves punk and hardcore music, Brazilian Jiu Jitsu, and sustainability.

Seven months ago, our engineers said letting an AI approve pull requests was too risky. They were right, so we didn’t start there.

Instead, we built a bot for code reviews and capped it at 200-line PRs, watched every decision, and turned every mistake it made into a rule or a lint check. Then we raised the limit.

This talk covers that loop: small experiments, proof instead of promises, and a system that gets a little better every week.

You’ll leave with a playbook you can start on Monday, whatever tools your company gives you.