Simon Kubica: Vibe coding for product leaders — from PRDs to pull requests in minutes

What the video covers

Simon Kubica, co-founder and CEO of Alloy and former product leader at Atlassian (Jira, Forge), delivered this talk at ProductCon in February 2026. The talk argues that the PM role is shifting from planning to building, and demonstrates what that shift looks like using tools available today. Kubica draws a line through three historical phases: a legacy era of whiteboarding and long development cycles; a “vibe coding” era of visual app builders that offered speed but remained disconnected from real product code; and the current builder era, in which AI agents work directly on live codebases and generate pull requests ready for engineering review.

Who it’s for

Product managers and product leaders at any level who work alongside engineering teams and want to reduce the feedback loop between idea and implementation. The talk is most applicable for PMs at companies using cloud-based development environments and modern SaaS tool stacks. It is less applicable for teams in heavily regulated environments or those with strict codebase access policies.

Key takeaways

1. Agents running overnight can replace multi-week planning cycles for defined features. Kubica demonstrates using Claude Opus 4.6 on 15+ hour asynchronous runs to develop features based on specifications, with the output being a pull request rather than a document. This shifts the PM’s role from communicating requirements to engineering toward reviewing what was built and deciding whether it matches the intent.

2. Success depends on context, not just instructions. Three prerequisites make this work: the AI must have access to the actual product codebase and design system so it can generate production-ready output, it must be connected to the team’s SaaS tools (Jira, Slack, Snowflake, Amplitude) so it has operational context, and its output must live in shared cloud environments rather than local machines so the whole team can see and test it without extra steps.

3. Customer calls can trigger builds in real time. Kubica demonstrates a workflow where a PM captures a feature request during a customer call, drops it into a shared environment, and has a functional prototype ready by the next morning. Feedback loops that previously spanned sprint planning cycles can compress to hours when the agent has enough context to act autonomously.

4. Roadmaps can become interactive rather than static. Instead of presenting a list of planned features as text or slides, product managers can share live prototypes of each initiative. Stakeholders interact with what is being proposed rather than reading descriptions of it, which changes how prioritization conversations go and what questions get asked early.

Worth watching if…

You are a product manager or product lead who wants to understand what AI-native product management looks like in practice rather than in general terms — or if you are evaluating whether to restructure your team’s development workflow around AI agents and want to see a specific demonstration of how that works end to end.