Surviving the AI-Native Era with Sander Hoogendoorn

This text was generated using AI and might contain mistakes. Found a mistake? Edit at GitHub

Key Takeaways

Trunk-based development and continuous deployment enable drastically shorter feedback loops (15 minutes to production), which fundamentally changes how teams work and interact with the business.
AI doesn’t replace the need for developers but enables them to build things previously considered too expensive or time-consuming, like custom UI components and internal tools.
The “product engineer” model eliminates intermediaries between developers and business stakeholders, enabling direct communication and collaborative decision-making through a transparent voting system.
SaaS providers and tool suppliers face declining markets as companies increasingly build custom solutions in-house using AI to replace expensive off-the-shelf platforms and services.
Proper architectural guardrails, automated testing, and standardized patterns allow teams to maintain quality and scale without significantly increasing headcount.
The specification language for software has shifted from code to natural language (English/Dutch), enabling non-technical people to create prototypes that engineers then make fit for production.

How does trunk-based development relate to AI-native development, and why is it more effective than pull request workflows?
What is the product engineer concept, and how does it differ from traditional software development organizational structures?
Will AI decrease the total number of developers needed in the industry, or will new possibilities absorb the productivity gains?
How should enterprises balance enabling business users to build their own software while maintaining production-readiness and architectural consistency?
What distinguishes professional software engineers from business users who can now prototype software using AI tools?
Will roles like product owner, scrum master, and project manager become obsolete in an AI-native era?

Trunk-based development: A version control strategy where developers integrate code into a single main branch frequently, minimizing long-lived feature branches and merge conflicts.
Continuous deployment: An automated practice that deploys every code change that passes testing directly to production without manual approval steps.
Product engineer: A developer who works directly with business stakeholders to understand requirements, build solutions, and ensure they work in production, eliminating traditional intermediary roles like product managers or project managers.
Domain-driven design (DDD): An architectural approach that structures software around the core business domain and uses a shared language between technical and business teams to model complex problems.
Microservices architecture: A system design pattern where applications are composed of small, independently deployable services that communicate through APIs, each potentially managed by different teams.
Agentic coding: A development approach where AI tools work autonomously or semi-autonomously to generate code based on natural language specifications and can refactor or modify existing codebases.

Claude: An AI language model that Sander identifies as a significant game-changer for code generation capabilities compared to earlier models.
Model Context Protocol (MCP): A protocol mentioned as part of early questions about integrating AI tools into development workflows.
Ant Design: An open-source UI framework from Alibaba that provides pre-built components, which Sander’s team partially replaced with custom solutions using AI assistance.
Vercel: A cloud deployment platform mentioned as an example of easy but unsecured deployment practices that enterprise developers must improve upon.
Jira and Confluence: Atlassian tools for project tracking and documentation that Sander’s team is considering replacing with custom-built alternatives.
Voucherify: A SaaS platform for managing coupons and promotional campaigns that Sander’s team evaluated for potential in-house replacement.